May 19 2023
Leave a comment
Artificial Intelligence

The state of the art in using AI to create images

By Eric Picard

This article is a discussion of the current state of generative image AI software, and how simple prompts can lead to beautiful images, but trying to get what you’re looking for exactly, and controlling for that is quite involved, and requires some tricks to really manage the AI’s interpretations.

Note: For this article, I’m using MidJourney Version 5.1.  You will see in the text prompts some commands that are unique to MidJourney, the most obvious being –v 5.1 which is the command telling the MidJourney AI which version to use.  5.1 is the newest model supported by MidJourney as of this day (Friday, May 19th, 2023).  There are lots of other generative AI tools out there, but personally I’m finding that MidJourney gives me the results I prefer over the others. Also, some of the prompts have been edited for clarity and consistency, but not in a way that affects the output. None of the images have been edited.

The idea I have for this project is to create an idealized image of a couple in their forties, standing in a New England field, arm in arm, with a colonial house in the background. The vision I have for this is reminiscent of an Irish Spring or Old Spice commercial from the 1970s, but set today. So I’m going to be including things in the prompt that aim the AI at recreating some of the sense of this from back in that time.  Here is what I would write as a prompt as a starting point if I were just doing this for myself:

“a photograph of a couple posing for the camera holding each other, they both wear thick white cable knit sweaters and are in their mid-forties. She is a beautiful, tall and willowy woman, he is clean shaven, rugged and handsome, they stand in a new england field with long grass and brightly colored wildflowers, in autumn, they look towards the camera, there are fieldstone walls in the background and a row of trees at the back of the field showing fall colors in red and yellow, and a small colonial house visible in the distance. There is depth in the photograph, with the house being positioned off in the distance, with rolling hills in between the couple and the house.”

I have an idea that this would provide me with decent results just off of a prompt, but I know well enough that I likely would need some image prompts, as well as text to get to my ideal end result. Including some visual instructions to the AI that would make it likely to include all the elements I care about. Also, I know roughly what I’m looking for in the way the characters look. I’d love the man to look like James Purefoy in his role in Fisherman’s Friends, but more clean shaven, maybe just stubble.  And the woman in my mind looks like Rachel McAdams dressed casually, or maybe Kate Mara with her aged makeup from the new show she’s in “the Class of 09”.

But rather than taking all my accumulated knowledge of how to write prompts, I’m going to start out simple, because you’ll see that the first set of images that MidJourney are quite beautiful, but don’t meet my initial vision.  

Starting with a simple prompt gets me this group of four images:

1:20 PM

a man and a woman stand in a field with long grass and wildflowers –v 5.1

Well – the field looks kind of like what I wanted, but none of the other background components are there, and the two figures are nothing like what I want.  I also know that trying to tune the whole image with two figures gets complicated, so I’m going to retrench and just start with a single figure, and I’ll use the man first.  I’ll start tuning my prompts until I start getting closer to what I want.

[1:21 PM]

a man stands in a field with long grass and wildflowers –v 5.1 –  

This is a good starting point, let me start tuning the man to get closer to my vision:

[1:23 PM]

a photograph of a tall man with broad shoulders wearing jeans and a sweater stands in a new england field with long grass and wildflowers in autumn, he looks toward the camera –v 5.1 

Okay – not quite what I’m looking for, but we’re getting there. Let me tune the man and the setting he’s in a bit:

[1:25 PM]

a photograph of a handsome and rugged tall man with broad shoulders, wearing jeans and a thick cable knit white sweater stands in a new england field with long grass and wildflowers of many colors in autumn, he looks towards the camera, there are stone walls in the background and a row of trees and a small house visible in the distance. –v 5.1 –  

That’s better but he’s not quite right, and I want more depth in the image. 

 — Yesterday at 1:28 PM

a photograph of a handsome and rugged tall man with broad shoulders who looks like a James Purefoy, wearing jeans and a thick cable knit white sweater stands in a new england field with long grass and wildflowers of many vibrant colors in autumn, he looks towards the camera, there are stone walls in the background and a row of trees showing fall colors in red and yellow, and a small colonial house visible in the distance. –v 5.1 

Okay, not really better. Yes, one of those images looks like James Purefoy, but I lost all the other elements that I love. Sometimes focusing too much of the AI’s attention on one element loses you the rest. Also, the aspect ratio of the image isn’t very cinematic, so I’m going to set the aspect ratio going forward to a 2:1 ratio using the –ar command: 

[1:30 PM]

a photograph of a handsome and rugged tall man with broad shoulders, wearing jeans and a thick cable knit white sweater stands in a new england field with long grass and wildflowers of many vibrant colors in autumn, he looks towards the camera, there are stone walls in the background and a row of trees showing fall colors in red and yellow, and a small colonial house visible in the distance. –ar 2:1 –v 5.1 –  

Okay – this is getting much closer, but he’s too young. I want to tune his age:

[1:31 PM]

a photograph of a handsome and rugged tall man in his forties with broad shoulders, wearing jeans and a thick white cable knit sweater stands in a new england field with long grass and wildflowers of many vibrant colors in autumn, he looks towards the camera, there are stone walls in the background and a row of trees showing fall colors in red and yellow, and a small colonial house visible in the distance. –ar 2:1 –v 5.1 –  

Okay, a bit more age specific, but I don’t have my white cable knit sweater, and I see that the AI is ignoring my request for jeans. Maybe if I remove jeans, the sweater will resolve?

[1:34 PM]

a photograph of a handsome and rugged tall man in a thick white cable knit sweater in his forties with broad shoulders, stands in a new england field with long grass and brightly colored wildflowers, in autumn, he looks towards the camera, there are stone walls in the background and a row of trees showing fall colors in red and yellow, and a small colonial house visible in the distance. –ar 2:1 –v 5.1 –  

That didn’t really help.  Let’s try specifying the color of the house and see if that helps. Also, these guys are all getting beards, so let’s clean that up:

[1:35 PM]

a photograph of a handsome and rugged clean shaven tall man in a thick white cable knit sweater in his forties with broad shoulders, stands in a new england field with long grass and brightly colored wildflowers, in autumn, he looks towards the camera, there are stone walls in the background and a row of trees showing fall colors in red and yellow, and a small white colonial house visible in the distance. –ar 2:1 –v 5.1 –  

Alrighty, the clean shaven prompt helped, but the white house request isn’t helping. And I still don’t have a white cable knit sweater.  Let’s see what happens when I include the “Stylize” command. This lets the AI be more creative in how it executes. The Stylize command also has a range of settings from 1 – 1000. I’ll put it on the highest setting to see what the difference is:

1:37 PM

a photograph of a handsome and rugged clean shaven tall man in a thick white cable knit sweater in his forties with broad shoulders, stands in a new england field with long grass and brightly colored wildflowers, in autumn, he looks towards the camera, there are stone walls in the background and a row of trees showing fall colors in red and yellow, and a small colonial house visible in the distance. –ar 2:1 –stylize 1000 –v 5.1 –  

Well, you can see that the AI was a bit more creative in the layout, but it didn’t get us what I want. Now it’s time to get really aggressive with the background. I’m going to put in some images to illustrate aspects of what I like.  I’ll use these three images for all the rest of the renderings, but here they are for your review:

This really helps the AI see that I want depth in the image, also what I’m looking for with a line of trees in the distance. It doesn’t really help with the wildflowers though, nor with the stone walls. So I’ll add in more images:

Let’s see how that helped, keeping Stylize turned on:

1:45 PM

https://s.mj.run/6oUsxj-acqs https://s.mj.run/M33VDlt9f2k https://s.mj.run/dKDRmKwbioI a photograph of a handsome and rugged clean shaven tall man in a thick white cable knit sweater in his forties with broad shoulders, stands in a new england field with long grass and brightly colored wildflowers, in autumn, he looks towards the camera, there are stone walls in the background and a row of trees showing fall colors in red and yellow, and a small colonial house visible in the distance. –ar 2:1 –stylize 1000 –v 5.1 –  

That was much better as far as putting depth and the wildflowers and the walls in – but let’s see if it works better with Stylize off:

[1:46 PM]

https://s.mj.run/6oUsxj-acqs https://s.mj.run/M33VDlt9f2k https://s.mj.run/dKDRmKwbioI a photograph of a handsome and rugged clean shaven tall man in a thick white cable knit sweater in his forties with broad shoulders, stands in a new england field with long grass and brightly colored wildflowers, in autumn, he looks towards the camera, there are stone walls in the background and a row of trees showing fall colors in red and yellow, and a small colonial house visible in the distance. –ar 2:1 –v 5.1 –  

Okay, so we’re getting somewhere.  I like the background quite a lot, although the figure is getting buried a bit.  Let me try this same set of images with a woman instead.

[1:48 PM]

https://s.mj.run/6oUsxj-acqs https://s.mj.run/M33VDlt9f2k https://s.mj.run/dKDRmKwbioI a photograph of a beautiful and willowy woman in a thick white cable knit sweater in her forties, stands in a new england field with long grass and brightly colored wildflowers, in autumn, he looks towards the camera, there are stone walls in the background and a row of trees showing fall colors in red and yellow, and a small colonial house visible in the distance. –ar 2:1 –v 5.1 –  

That’s good – I’m seeing a consistent treatment of the setting, but a lot of play with the figure. Let’s see what Stylize does.

[1:50 PM]

https://s.mj.run/6oUsxj-acqs https://s.mj.run/M33VDlt9f2k https://s.mj.run/dKDRmKwbioI a photograph of a beautiful and willowy woman in a thick white cable knit sweater in her forties, stands in a new england field with long grass and brightly colored wildflowers, in autumn, he looks towards the camera, there are stone walls in the background and a row of trees showing fall colors in red and yellow, and a small colonial house visible in the distance. –ar 2:1 –stylize 1000 –v 5.1 –  

That didn’t seem to make a lot of difference.  So now I’m going to start playing with the figure and see if I can tune it towards what I want.  Let’s start with a general figure that’s the right age and wearing the sweater I want:

 — Yesterday at 1:52 PM

photographic portrait of a Handsome Man, cleanshaven and rugged looking, in a thick white cable knit sweater –v 5.1 

Well, that’s not James Purefoy, but I think the 3rd image looks pretty good. I’ll render that out and look at it larger:

1:55 PM

photographic portrait of a Handsome Man, cleanshaven and rugged looking, in a thick white cable knit sweater –v 5.1 – Image #3  

Now that I have an image that gets more of the visual information to the AI, I’ll drop that into the first slot of my prompt:

2:00 PM

https://s.mj.run/ePXrbTX226c https://s.mj.run/6oUsxj-acqs https://s.mj.run/M33VDlt9f2k https://s.mj.run/dKDRmKwbioI a photograph of a handsome and rugged clean shaven tall man in a thick white cable knit sweater in his forties with broad shoulders, stands in a new england field with long grass and brightly colored wildflowers, in autumn, he looks towards the camera, there are stone walls in the background and a row of trees showing fall colors in red and yellow, and a small colonial house visible in the distance. –ar 2:1 –v 5.1

And this is much better. Really close to what I’m looking for.  Now I just need to find a woman to help tune the image.  Here’s my first try on that.

[2:02 PM]

beautiful and willowy woman in a thick white cable knit sweater in her forties –v 5.1   

I find that MidJourney struggles a bit with age, particularly women.  I’d say these women look older than mid-forties. More like fifties or even early sixties.  But for now, it should at least set the tone.

 2:03 PM

beautiful and willowy woman in a thick white cable knit sweater in her forties –v 5.1 – Image #3  

[2:05 PM]

https://s.mj.run/DDmJPXMNGS0 https://s.mj.run/6oUsxj-acqs https://s.mj.run/M33VDlt9f2k https://s.mj.run/dKDRmKwbioI a photograph of a beautiful and willowy woman in a thick white cable knit sweater in her forties, stands in a new england field with long grass and brightly colored wildflowers, in autumn, he looks towards the camera, there are stone walls in the background and a row of trees showing fall colors in red and yellow, and a small colonial house visible in the distance. –ar 2:1 –v 5.1 –  

Great – this gets us pretty close to what I’m looking for, but let me try both the male and female version of these with Stylize maximized and see if it helps…

[2:09 PM]

https://s.mj.run/ePXrbTX226c https://s.mj.run/6oUsxj-acqs https://s.mj.run/M33VDlt9f2k https://s.mj.run/dKDRmKwbioI a photograph of a handsome and rugged clean shaven tall man in a thick white cable knit sweater in his forties with broad shoulders, stands in a new england field with long grass and brightly colored wildflowers, in autumn, he looks towards the camera, there are stone walls in the background and a row of trees showing fall colors in red and yellow, and a small colonial house visible in the distance. –ar 2:1 –stylize 1000 –v 5.1 –  

[2:10 PM]

https://s.mj.run/DDmJPXMNGS0 https://s.mj.run/6oUsxj-acqs https://s.mj.run/M33VDlt9f2k https://s.mj.run/dKDRmKwbioI a photograph of a beautiful and willowy woman in a thick white cable knit sweater in her forties, stands in a new england field with long grass and brightly colored wildflowers, in autumn, he looks towards the camera, there are stone walls in the background and a row of trees showing fall colors in red and yellow, and a small colonial house visible in the distance. –ar 2:1 –stylize 1000 –v 5.1 –  

I think that actually helped a bit, but not a ton. Now we’re ready to tune our image and text prompts to get closer to our final version.  I’m going to create a new prompt just to get the couple generated in a way that will help the AI.  As you’ll see in a moment, even with an explicit prompt, MidJourney really wants to blend two input images into a single image, especially with people.  So I know before I do this that I’m going to end up with a few different blends, but usually at least one image in the quad will follow the instructions. For this I’ve taken my two individual images of the man and woman, and dropped them in just to see what happens:

2:14 PM

https://s.mj.run/ePXrbTX226c https://s.mj.run/DDmJPXMNGS0 photographic portrait of a couple in their forties arm in arm, he is clean shaven rugged and handsome in a thick white cable knit sweater, she is tall and willowy wearing a fall colored sweater –v 5.1 –  

Okay – that was three blended single figures, and one that matched more what I was looking for. So I’ll take that one and see what happens.

 2:15 PM

https://s.mj.run/ePXrbTX226c https://s.mj.run/DDmJPXMNGS0 photographic portrait of a couple in their forties arm in arm, he is clean shaven rugged and handsome in a thick white cable knit sweater, she is tall and willowy in her late thirties wearing a fall colored sweater –v 5.1 – Image #2  

Just so that I can tune this, let’s try just a text prompt and see what we get with no reference images, because this isn’t 100% what I’m looking for:

[2:16 PM]

photographic portrait of a couple in their forties arm in arm, he is clean shaven rugged and handsome in a thick white cable knit sweater, she is tall and willowy wearing a fall colored sweater –v 5.1 –  

Remarkably, I think not using reference images gave me a couple I could use here that is much closer to what I was looking for – with the third image.  

2:17 PM

photographic portrait of a couple in their forties arm in arm, he is clean shaven rugged and handsome in a thick white cable knit sweater, she is tall and willowy in her late thirties wearing a fall colored sweater –v 5.1 – Image #3  

[2:20 PM]

https://s.mj.run/F9kR5x8XeRM https://s.mj.run/6oUsxj-acqs https://s.mj.run/M33VDlt9f2k https://s.mj.run/dKDRmKwbioI a photograph of a couple in their forties wearing thick white cable knit sweaters, posing for the camera holding each other, she is a beautiful, tall and willowy woman , he is tall clean shaven, rugged and handsome, they stand in a new england field with long grass and brightly colored wildflowers, in autumn, they look towards the camera, there are stone walls in the background and a row of trees showing fall colors in red and yellow, and a small colonial house visible in the distance.  –ar 2:1 –v 5.1 –  

Now we’re cooking with fire.  I really like the first image, but there’s no house in it.  Let’s see if Stylize helps.

[2:26 PM]

https://s.mj.run/F9kR5x8XeRM https://s.mj.run/6oUsxj-acqs https://s.mj.run/M33VDlt9f2k https://s.mj.run/dKDRmKwbioI a photograph of a couple in their forties wearing thick white cable knit sweaters, posing for the camera holding each other, she is a beautiful, tall and willowy woman , he is tall clean shaven, rugged and handsome, they stand in a new england field with long grass and brightly colored wildflowers, in autumn, they look towards the camera, there are stone walls in the background and a row of trees showing fall colors in red and yellow, and a small colonial house visible in the distance.  –ar 2:1 –stylize 1000 –v 5.1 –  

Okay – that second image in the upper right is just about perfect.  I’d like it if he was wearing a sweater with a bit more texture and thickness.  And the AI has ignored my original request to have her in a cable knit sweater too.  But I like it even better.  So after all that, we have our final image:

2:27 PM

https://s.mj.run/F9kR5x8XeRM https://s.mj.run/6oUsxj-acqs https://s.mj.run/M33VDlt9f2k https://s.mj.run/dKDRmKwbioI a photograph of a couple in their forties wearing thick white cable knit sweaters, posing for the camera holding each other, she is a beautiful, tall and willowy woman , he is tall clean shaven, rugged and handsome, they stand in a new england field with long grass and brightly colored wildflowers, in autumn, they look towards the camera, there are stone walls in the background and a row of trees showing fall colors in red and yellow, and a small colonial house visible in the distance. –ar 2:1 –stylize 1000 –v 5.1 – Image #2

As you can see, this whole process from start to finish took just over an hour, and required me to think a lot about what I wanted, and to be willing to iterate quickly.  I also could have shortened that experience by starting with my very first prompt and iterating from there.  If I go back to that very first prompt, you’ll see that it gives us something good – but doesn’t really get me to what I had envisioned.

“a photograph of a couple posing for the camera holding each other, they both wear thick white cable knit sweaters and are in their mid-forties. She is a beautiful, tall and willowy woman, he is clean shaven, rugged and handsome, they stand in a new england field with long grass and brightly colored wildflowers, in autumn, they look towards the camera, there are fieldstone walls in the background and a row of trees at the back of the field showing fall colors in red and yellow, and a small colonial house visible in the distance. There is depth in the photograph, with the house being positioned off in the distance, with rolling hills in between the couple and the house. –ar 2:1 –v 5.1”

These are all very nice images, although we see MidJourney having its age issue again, and not listening well to the clean shaven input.  We’ve also lost a lot of what we gained from using the reference images, so I’ll drop those back in:

https://s.mj.run/F9kR5x8XeRM https://s.mj.run/6oUsxj-acqs https://s.mj.run/M33VDlt9f2k https://s.mj.run/dKDRmKwbioI a photograph of a couple posing for the camera holding each other, they both wear thick white cable knit sweaters and are in their mid-forties. She is a beautiful, tall and willowy woman, he is clean shaven, rugged and handsome, they stand in a new england field with long grass and brightly colored wildflowers, in autumn, they look towards the camera, there are fieldstone walls in the background and a row of trees at the back of the field showing fall colors in red and yellow, and a small colonial house visible in the distance. There is depth in the photograph, with the house being positioned off in the distance, with rolling hills in between the couple and the house. –ar 2:1 –v 5.1 

This is much better, and I could iterate on this to really get to an image that is as good as the one we ended up with from the longer process without as much work.

This is where the state of play is today in Generative AI for images.  As a product person, I have a lot of ideas for what needs to happen from a design tools perspective to really supercharge the process.

For instance, I should be able to spend a lot of time generating a single “entity” like “Man in white cable knit sweater” and use that single entity over and over – without having it change each time.  I should be able to generate a landscape, get it perfect, and then drop other entities into it. Today that isn’t possible, but you can imagine that this would be a game changer.  

I’ve spent a lot of my career building design tools and working in advertising, so I know pretty intuitively what a designer needs in order to live up to the requirements of working with clients.  The need to get creative approval on the specific characters (the exact face, the exact sweater, the exact color sweater) are all things that the client would want to approve.  Today with generative AI each time you render that prompt, even with really good images input into the model, it’s very hard to get consistent results.  

But you probably can now really get a sense of what the future holds.  AI powered design tools are going to change everything.  And a lot of careers are going to morph over the next few years.

Tagged ,
May 17 2023
1 Comment
Artificial Intelligence, Predictions, Software Development

How AI will actually impact software development

By Eric Picard

As a product person, I am constantly intrigued by emerging technologies and how they shape the future. The true game-changer of our age, possibly in all of history, is Artificial Intelligence (AI). There’s a lot of hype, and when I hear people talk about AI right now, it’s often discussed as if it’s a mature offering.  Nothing could be further from the truth, we’re in AI’s earliest stages. But the potential is vast. 

I’ve had the great fortune to work with some of the smartest engineers in the world, and in my recent conversations with them I’ve gained some insights into what lies ahead. These engineers, often referred to as 10X engineers (a few are even more than 10X) due to their exceptional productivity, are in awe of what they’re seeing. I’ve collaborated with these engineers in the past, witnessing their ability to tackle complex technical problems, invent new technologies and products, sometimes on a daily basis. 

I’ve been hearing remarkable stories from this cohort, some of whom are among the best living engineers. Overwhelmingly I’m hearing that they have experienced an astonishing 100X increase in productivity by utilizing ChatGPT 4 and other open-source Large Language Models. Some have built working prototypes where AI-powered bots with specific skill sets form functional teams. These bots collaborate, architect solutions, and even write code based on detailed business problem prompts. Although some fine-tuning is necessary, the progress is truly remarkable, and is getting better very fast.

Conversational Software Development: A Paradigm Shift

Spending a significant amount of time immersed in ChatGPT and Google’s Bard, I am convinced that we are on the verge of a monumental shift in software development. Conversational software development powered by AI is set to make agile methodologies seem quaint in comparison. This transformation will revolutionize the industry, enabling projects to be completed with far fewer engineers than before.

Before you misunderstand, let me dispel the notion that this reduces the value of engineers. Over the past three decades, there has been a severe shortage of engineering talent. Far from diminishing the need for engineers, AI will exponentially increase the amount of software being written with the same number of engineers. It will propel us into a world where agile sprints and extensive planning will be replaced by direct transitions from concept to functional prototypes.

Picture a scenario reminiscent of the Iron Man movies, where Tony Stark converses with Jarvis, his AI assistant, rapidly manufacturing and discarding prototypes until the desired outcome is achieved. Every prototype becomes a complete product and every line of code will become disposable, allowing AIs to iteratively build new versions with seamlessly integrated upgrades. This fundamentally changes our conceptions of technology development and product lifecycle management. Far from being unnecessary, engineers will become even more valuable, with the smartest among them assuming the role of architects of the AI and the development processes, shaping the infrastructure.

Evolving Roles and New Opportunities

In this transformed landscape, certain roles will take center stage. Product Management will become even more critical, with many engineers transitioning to product-focused positions. Crafting executable prompts that AI models can understand will be a new skill set for product-oriented individuals, who will define what needs to be built.

Collaborating closely with Product Managers, software architects will play a pivotal role. Their expertise will ensure that architectural decisions align with product goals and requirements, providing guidance while engineers focus on implementation.

While the roles of Product Marketing, Sales Enablement, and Marketing Communications will remain crucial, they will need to adapt to the surge of software products flooding the market. Differentiating and creating demand through effective marketing will become even more vital in this highly competitive landscape.

About ten years ago I heard Quentin George (who was at advertising agency holding company IPG at the time) give a presentation where he said something like, “There is a massive overabundance of manufacturing capacity in the world, and too many products are rolled out every year. Most of this output goes to the landfill. The only way to compete in a manufacturing economy with such overabundance of capacity is by using marketing to differentiate and create demand for your product.”  

This same set of problems will exist for the first time in software development. When there is not a shortage of engineering capacity, and we 100X the throughput of engineers, we’re going to have a lot more software exploding into the world. Marketing becomes critical in that scenario – in ways it simply wasn’t in the past, because there wasn’t a ton of competition.  Consumer software development is about to explode in volume.

The AI revolution is poised to transform software development by exponentially increasing productivity, automating processes, and moving from a long development cycle to a rapid development and iteration cycle. It will be faster to simply write the code and build the product than to design, mockup, write specifications, drive alignment, then go through multiple development sprints. We’ll simply build working products, and if they’re not correct, we’ll throw them out and build better versions.

Jan 10 2023
1 Comment
Ad Ecosystem, Economics, Online Media, Predictions, Uncategorized

5 Year Predictions – January 2023

Once every few years I like to write an article predicting what will happen in the future. Over the years I’ve had a pretty good track record of getting things right.  The world is shifting and moving a lot right now, but I believe that the future is bright.  Here’s how I think about the next five years, and beyond, through the lens of Ad Tech, Consumer Technologies, Media and Advertising.

  1. 3rd Party Cookies won’t go away, but they will slowly be rendered non-usable as persistent IDs

3rd party cookies, which have been a commonly used tool in the ad tech industry, will not completely disappear but will instead become increasingly less useful as persistent IDs. Google, for example, will not shut off 3rd party cookies in Chrome, but will instead make them less usable for persistent IDs over time. This gradual decline in functionality is expected to take place over a period of five to ten years, and by the end of this time frame, we will likely see the value of 3rd party cookies in the ad tech space significantly decrease. In five years, we will already be on this trajectory towards the obsolescence of 3rd party cookies as persistent IDs.

  1. New approaches to targeting inventory that are privacy-centric will arrive at scale

As the ad tech industry shifts away from 3rd party cookies as persistent IDs, new approaches to targeting inventory that prioritize privacy will become increasingly prevalent. These new methods will be built on technologies such as cohorts and will make use of panels of users that are statistically relevant. This will allow advertisers to not only target the audiences they care about but also more effectively attribute their advertising spend to various outcomes. 

The approaches currently being developed, including techniques such as embeddings and deep learning, will greatly surpass the current “brute force” methods used in ad tech and will lead to a move away from surveillance-based approaches towards those that prioritize privacy. Additionally, publisher and advertiser first-party data will be used to feed these privacy-centric models. The technology and techniques to match supply-side and demand-side data already exist, and this process will become increasingly easy, privacy-conscious, and available at scale. 

This will lead to a more equitable understanding of customer behavior and reduce the information imbalance that has favored the buy side in recent years. The seed audiences that act as panels for ML models will lead to more equilibrium of understanding customer behavior and reduce the information imbalance that has grown over the last decade in favor of the buy side.

  1. The lines between Buy and Sell Side ad technologies will blur

The lines between buy-side and sell-side ad technologies are becoming increasingly blurred. Companies like The Trade Desk are beginning to integrate directly with publishers, bypassing the SSP and exchange infrastructure. In response, SSPs and exchanges are starting to offer buying platforms, allowing buyers to bypass DSPs. This trend will continue for a few years, reach a peak, and then ultimately collapse in on itself. 

This is because DSPs are designed to lower competition over inventory and keep prices as low as possible, which is in line with their role as representatives of the buy-side. However, their algorithms are designed from a buyer’s perspective, and publishers will be wary of these direct paths, resulting in a decrease in yield. 

Exchanges and SSPs have mostly focused on liquidity and passing inventory through to DSPs at the lowest cost possible, while publishers have continued to lose power in the struggle between the buy-side and sell-side of the market. However, the pendulum will ultimately swing back towards equilibrium, and publishers will regain more control over data and measurement. Companies will invest heavily in ways to increase publisher yield and the market will balance out again.

  1. Web 3 Technology will iterate beyond just Cryptocurrency 

Web 3 technology is evolving and shifting beyond cryptocurrency, towards solutions that support distributed identity and group collaboration. This will have a significant impact on advertising in several ways. Imagine a world where users have full control over their identity and data, and only share relevant information with the companies they choose to interact with, through mechanisms that obscure unnecessary information. Healthcare and finance industries already use some techniques for doing this at scale, and combining these techniques with approaches used in the Web3/crypto space can open up new possibilities. For example, a digital wallet that contains all the important information about an individual’s life, such as healthcare, financial, education, employment, real estate, municipal and government information and automatically shares only relevant information with companies and organizations.

Users could easily opt-in to being part of a brand’s community, which would merge CRM, CDP, Ad Serving, and Social Media. This would mean that users get special perks from that brand, including the ability to get special offers, customized products, early access, etc. Brands could reach out to users and ask for their opinions on products and reward them for their participation. Users could “stake” their interest in a new product or feature and in return get early access, similar to an Indiegogo campaign, but for major brand interactions. Users could also vote on product changes or feature prioritization based on their staking, and the staking could be based on a points system based on their loyalty.

For example, if you have owned five BMWs over the last 20 years, and you are a known high-value customer, you could participate in a user group of other high-value customers and apply your influence to get special options for your next car, or maybe even for mainstream features in all models. Maybe BMW would offer a limited-edition model just for that group of customers, or a special badge. Or maybe you and others have strong opinions about the placement of cup holders, and could influence a change in future models. The “staking” in this case could be the fact that you have already bought several BMWs, and you currently own one or more.

These concepts like Staking are common in the Web3 and Crypto space but haven’t yet gone mainstream. But in the next five years, we are likely to see more and more of these concepts being integrated in the mainstream industry, even if the behind-the-scenes mechanism is obscured from the customers.

  1. Retail Media Marketplaces will grow and expand. 

Retail media marketplaces are expected to grow and expand in the coming years. For big retailers like Amazon, Walmart, and Target, this represents an opportunity for additional revenue at higher margins. These networks have already expanded into grocery chains, and even to boutique e-commerce and retailers. They could expand even further beyond the virtual world and into the physical space between bricks-and-mortar stores.

The growth of these retail media marketplaces is due in part to the evolution of the old “coop-dollar” systems that have been in place for decades into something much more advanced. Brands can now pay for product placement in the search results for similar products. When combined with e-commerce experiences, this leads to better outcomes for all parties involved – brands, consumers, and e-commerce retailers. The margins on these media businesses are significantly higher compared to other parts of retailers’ businesses, which is why it is expected to proliferate.

Retailers have a direct consumer relationship, pure first-party data about the customer, and the positioning of these media units is almost perfectly located between the moment of purchase consideration and the purchase itself. This means brands will be willing to spend money on this “must-buy” piece of media. Additionally, bundling of virtual shelf placement with in-store environments will make this buy even stickier over time. If brands want to get good shelf positions, end-caps, and other in-store benefits for their products, they will need to also pay for placement in the virtual space. Ultimately, these will blur and blend and package together, but it is likely further out than five years.

  1. Social Networking will evolve to something else altogether. 

Social networking is expected to evolve into something else altogether, with everything tied back through the social graph. This includes commerce, communications, education, search, and more. The social graph maps the connections between people and their interests, and platforms like Facebook understand who you know and the flow of information between you and your connections, as well as their interests and sentiments on various topics.

If the social graph were to become open, meaning it is no longer a walled garden, and your identity and the social graph extends beyond people to companies, products, brands, media, music, film, etc., and where you, the human, are in control, and it’s easy to manage, there would be significant opportunities for growth and change. Social graphs would connect everything, and the consumer would be in charge. Applications built on top of these open social graphs would be different from anything we have seen so far.

Facebook has already become Meta, and they’re trying to own the metaverse. But even without virtual reality, the social graph overlaid across everything would be transformative. It could lead to collaboration between ephemeral and permanent groups of people to do things together. For example, it would be easy to organize a friend group to buy out a restaurant for an evening party, find 800 people in the greater Boston area who also love the New England Patriots and want to have a meet and greet with the team, or have dinner at a local restaurant with a special menu with ten of your closest friends.

But this is just the tip of the iceberg. Connecting the social graph to everything else will change the world. And if identity is solved, so we know you’re not a Russian Bot, things will only get better.

  1. Artificial Intelligence will change everything.  

Artificial Intelligence (AI) is expected to change everything in the coming years. We are already familiar with AI-powered solutions such as filters in Instagram or “lenses” in Snapchat, and predictive text to help with text messaging and correcting grammatical errors in documents. But these are just the beginning of a trend that is now starting to take off.

One example of this is ChatGPT, a new chatbot by OpenAI that complements their DallE offerings. ChatGPT enables the creation of very complex written content that can be indistinguishable from content created by humans. Some software developers are even using it to both bug-check and write code from scratch.

Similarly, AI image and video generators are on the cusp of making significant strides. MidJourney, Dall-E 2, and numerous other solutions can generate images in almost any style just by describing what one would like to see. The results are getting exponentially better on an ever-shortening curve.

While it’s important to note that this technology also brings ethical concerns such as copyright and originality, which need to be addressed, the gains will outweigh these concerns. Over the next few years, the art of combining human input with computer-generated output will be refined, and every single software tool used for writing, office work, finance, design, etc., will be transformed. Corporate users will have AIs trained just with their own datasets, such that trade secrets and non-public information can be incorporated into the AI engines. For creators, the initial concerns about artists having their work stolen by these AI engines will be replaced by new understandings of how artists can have their own AI, trained on their behalf, to supercharge and speed up the creation and generation of work.

For production artists and graphic designers, these AI tools will become a seamless and integral part of their workflow, allowing them to create and generate content faster and more efficiently. Musicians will also have access to similar tools that will allow them to compose, produce, and record music in new and innovative ways. The impact of AI in these fields will not only change the way we create and consume art, but it will also open up new possibilities for expression and creativity.

  1. The long game:  What big technology will sneak up on us and change all aspects of society?

    I’m going to say something that will sound boring:  Electricity.

When I do these predictions, I like to pick one long term trend and extrapolate even further out than 5 years.  The biggest trending technology I can think of is Electricity.  

Solar technology will continue to improve on a scale increase similar to Moore’s Law, which it has been meeting or beating for more than twenty years. Today the cost of solar power is about $0.08 per Kilowatt. If the costs keep dropping and the output keeps increasing on the same scale it has, electricity will become extremely cheap. 

You may recall how 20 years ago you paid a long distance fee for all phone calls except for local calls (just the town you lived in). Electricity will never be totally free, but similarly to how we now basically have free calling to anyone, anywhere, even video calling, we’re approaching a world where the cost of electricity is going to be so low, and the ability to create a distributed electrical grid and expand it everywhere will be so low, that the long term prediction should be for a very low cost and low or even zero emissions. Solar everywhere and incredibly cheap electricity will transform the way the world works eventually. 

Over the next five years, you should expect to see a lot more solar power implemented, on houses, on buildings, and even the beginnings of solar panels placed under fixed infrastructure like streets and parking lots. https://solarroadways.com/ 

Once that transition happens at scale, with free electricity nearly everywhere, you’ll see big shifts. There will be a convergence with other lower cost technologies like LEDs (Light Emitting Diodes) hitting their next generation, where laser diodes will become cheap enough to replace LEDs.  Lasers put out 1,000 times as much light as an LED, for only two-thirds as much energy.  When the laser diodes become cheap enough, and the power is almost free, we’ll see a revolution in lighting and therefore in video.  Effectively this means video everywhere, all the time. Streets made up of solar cells that have laser diodes mounted into the transparent high-strength glass surfaces so that the roads light up and animate.  Buildings covered in solar cells with laser diodes embedded in them, instant christmas lights, video on the side of buildings everywhere, and the ability to put lighted animated signage anywhere for nearly no cost. Streetscapes and cities will radically transform when this happens. 

And I’m bullish about carbon emissions because solar will be so cheap and the innovations on top of a newly formed, completely distributed solar grid are massive.

  1. And as always, my final prediction:  

Sometime in the next five years, some new technology nobody has even thought about, or a simple reinvention of an existing widely used technology, will come into existence and totally scramble things. Just like the iPhone was unexpected, just like the success of Social Media was unexpected, something new will appear. And once again it will change everything.

Feb 24 2021
Leave a comment
Ad Ecosystem, Ad Technology, Uncategorized

The 6th Wave of Advertising Technology: Privacy

By Eric Picard, Originally published on AdExchanger – Wednesday, February 24th, 2021

There’s a revolution happening in digital media, primarily driven by a new focus on privacy. Major players at the core of the digital ecosystem have decided that privacy is a core value, and have made fundamental changes that block many standard practices. This change is going to upend the industry as we know it, and offers huge opportunities for anyone in the right position to take advantage of it.

Let’s work our way from where we’ve been to where we are, and then talk about where we’re going.

Wave 1: In the beginning (1996-1998)

The first wave of ad tech was about establishing scalable ways to operate the digital advertising business. Someone had to figure out how to sell ads in advance of the campaign running, how to implement and operate campaigns, how to track delivery and how to bill customers. We saw the rise of ad servers, the creation of sales and ad operations tools and workflows and the invention of buy-side ad serving. And we saw significant growth.

Wave 2: Formats, Targeting, Tracking, Attribution 1.0 (1999-2001)

After the basics got sorted, we saw innovative work in rich media ad formats (things like interactive ads, video, audio, visual effects, over-the-page, expanding ads, etc.). My first startup, Bluestreak, developed many of these formats. Across the industry we saw significant innovation in targeting of ads. (User behavior was tracked and turned into audience segments, which could be sold.) And a new attribution discipline emerged to measure what happened after a person saw or clicked on an ad.

Wave 3: Remnant Monetization, Multi-Touch Attribution, Yield Optimization (2002-2006)

When the “dot-com” bubble burst in 2001, the average CPM of display ad inventory dropped from about $25 to about $0.50 in the course of a year. All the peripheral ad tech companies that had been charging ad-on fees for rich media and targeting began to struggle – that is, until they eventually realized they could sell directly to publishers as a way to drive yield. In the hunt for revenue at any cost, and as vast numbers of smart sales people got laid off, someone figured out that secondary and even tertiary ad marketplaces could be used to monetize every single impression at some price. This model was in some ways a mistake, because it further devalued inventory, which was already under price pressure. It took a long time for this wave to end, and in some ways it still hasn’t ended.

On the buy side, advertisers began to realize that “last touch” attribution was obscuring the real drivers of conversions, falsely rewarding some channels, specifically paid search. Sadly, some advertisers still use last-touch models.

Wave 4: The rise of programmatic (2007-2014)

A few really smart people realized that remnant marketplaces were evolving similarly to commodities and securities marketplaces. And they began building auction-based exchanges that sold inventory in much the same way paid search was sold.

This wave was incredibly powerful and it supercharged the industry. As we saw with the evolution of electronic exchanges in securities, the market moved away from daisy-chained tag-based auctions to real-time bidding (RTB). This extensible infrastructure also led to opportunities for nefarious actors to make money by fraudulently selling fake ads, defrauding advertisers and publishers of billions of dollars over many years. And a massive investment in the data infrastructure has led in many ways to a “surveillance state” that allows almost any company to track people’s behavior across the entire internet and build targeting segments that can be used to buy them as advertising.

Wave 5: Privileged Programmatic and Fraud Cleanup (2015-2020)

As it matured, programmatic advertising continued to walk in the footsteps of the securities exchanges. The largest ad buyers and sellers began to recreate privileged relationships inside the new RTB infrastructure. Examples include PMPs, “first look” mechanisms like header bidding and Prebid. It is now possible (but will take a while) to completely recreate all the ways ads have been sold historically on top of RTB infrastructure, but the eventual result will be a much more scalable and automated way of doing business.

Similarly, the massive and hidden problem of fraud was uncovered, and measures were taken to root it out. Industry efforts like Ads.txt and Sellers.json, and whole new companies and technologies for fraud detection and prevention, has set the industry on a path to solving this crisis. The result: a massive maturation of the ecosystem.

Wave 6: Privacy – Centric Advertising, New Format Innovation and Supply-Chain Optimization

Meanwhile the “surveillance state” we’ve found ourselves in has led to a huge backlash against third-party tracking that is upending the ecosystem again.

Over the last few years we’ve seen major initiatives by the technology industry to establish and enforce new privacy controls across all media. This trend is accelerating and broadening, and many of the mechanisms we’ve taken for granted in online advertising have been ruled privacy-unsafe, and are being phased out. Many companies in the space have doubled down on a commitment to these older tracking approaches, and are trying to find a path through that perpetuates them. I will hazard a prediction that this is not going to work.

The time is coming to an end when companies with no relationship to the consumer can track those consumers’ behavior across the internet and then sell that data. This evolution will strengthen companies that do have a direct (i.e. “first party”) consumer relationship, such as advertisers and publishers. It also is helping the largest incumbents like Facebook and Google, who have immense amounts of first-party data.

Technology providers will need to find ways to evolve their offerings such that they support the direct consumer relationships held by the advertiser and/or the publisher. This will mean in many cases either a completely new approach, or a set of innovations in how technology is integrated with the first-party companies’ infrastructure. The great thing about disruption is that it leads to new innovations.

Because the third-party data and tracking infrastructure is becoming less valuable, new ways to increase the value of ad opportunities will come to the foreground. Format innovation is back in the mix as a way to increase the value of inventory without breaching privacy protections. And the next wave of supply cleanup, after the war over fraud, will ensure supply chains are clean and optimized, with low-value suppliers shuffled out of existence.

Supply-chain optimization has been emerging as a focus area since around 2014, but now is becoming mainstream. The first round of supply-chain optimization was ‘brute force’ and ugly, but we’re now seeing intelligent and powerful supply-chain optimization enter the market, as well as industry initiatives like Sellers.json. These new technologies, initiatives and approaches are driving advertiser value and publisher yield significantly.

Tagged ,
Apr 03 2017
Leave a comment
Ad Ecosystem, Ad Technology, Economics, Marketers, Media Agencies, Online Media, Predictions, Publishers

The Fifth Wave Of Ad Tech: Privileged Programmatic

By Eric Picard (originally published on Adexchanger.com Friday, March 10th, 2017)

The first seven years of the programmatic revolution were driven by three major efforts.

It began with the creation and propagation of the massive new infrastructure needed to support real-time bidding. That was followed by the connection of all demand to all supply in the programmatic infrastructure. New ad products, formats and platforms then emerged, built on top of this new infrastructure.

This was a significant revolution – what I’ve called the third and fourth waves of ad technology. We’re now entering a fifth wave: privileged programmatic.

As the programmatic ecosystem matures, we’re seeing massive adoption of these new tools and technologies by the largest advertisers and media agencies now spending at scale. During the first seven years or so, many ad networks procured and resold media and some large marketer early adopters broke ground – many of which are now reaping the dividends.

But the very largest budgets are now coming into programmatic, and the game is changing. To illustrate the change, let’s talk about the historical evolution that the financial markets went through as they hit their maturity threshold during the rise of electronic trading.

Lessons From High-Speed Traders

In the highly recommended book “Dark Pools: The Rise of the Machine Traders and the Rigging of the US Stock Market,” by Scott Patterson, there is a clear narrative that will start to feel familiar to those working in programmatic media.

When the electronic markets were created, the early winners were typically hedge funds established and managed by the same humans who built the electronic market infrastructure. They knew that traders that responded the fastest to auctions could get significant advantage over other participants. Thus began the high-speed trading (HST) revolution. High-speed traders made millions of dollars a day on high-volume trades at very high speeds.

As the market matured, large traditional stock market players entered the electronic trading business and had their lunch eaten by the upstart high-speed traders. They found this to be unacceptable. The basic logic was, “If I’m spending billions of dollars a year on your electronic exchange, I need some privilege that gets me ahead of these little upstarts who have ‘know-how’ but are tiny players compared to me.”

The biggest players went to the exchanges and demanded privileged bidding mechanisms to allow them to win in the auction even if another player bid higher or bid first. They removed the advantage built in by the high-speed traders.

Nobody warned the HST companies. Within weeks in some cases, many simply went out of business. They had no idea what happened, but knew they suddenly weren’t winning in the auction. Eventually a few found out that the unpublished bid mechanisms that allowed them and the large brokerages to win in the auction had been uncovered and made available more broadly. But most of the damage was already done.

Privileged Programmatic

Privilege in an auction environment is not necessarily a bad thing. Much like the RTB exchanges in advertising, the electronic markets were seen as the great equalizers – fair unbiased auctions – but the reality is that the HST companies had their own type of advantage based on infrastructure knowledge. A real business argument can be made that buyers spending vast amounts of money should be able to negotiate for privilege with the sellers. That’s exactly what is happening in programmatic advertising.

Have you noticed that many of the biggest early players in programmatic have come upon hard times? Suddenly algorithms that were designed to provide advertisers with performance while still stripping off big dollars via an arbitrage model stopped working. Why?

Over the last few years we’ve seen the massive adoption of new privileged mechanisms in programmatic. Whether we discuss private marketplaces (PMPs), header bidding, first-look or programmatic guaranteed, they are predictable artifacts of the maturation of the programmatic marketplace. And don’t let any early knowledge you’ve gathered on these mechanisms create a false sense of comfort – PMPs from three years ago often look nothing like the configuration seen today. These mechanisms are not created equally.

For publishers, this maturation is very good news. Many large publishers viewed programmatic as a “rush to the bottom” in the early days and now see programmatic mechanisms bringing balance back to the marketplace.

Many publishers expressed frustration as programmatic created for the first time in digital media an information asymmetry that favored advertisers. Publishers had no idea why advertisers bought media from them over the open exchange, and now with these privileged mechanisms, the conversation has moved back to media buying and sales teams are empowered to negotiate and structure deals that drive customer value.

The hallmark of the first seven years of programmatic was a bottoms-up reinvention of buying based on data-driven decisioning and performance – and the biggest lever on performance was price of inventory. Early adopters were astounded to find their desired audiences for a low cost on the exchanges, even at the same publisher sites where they were simultaneously executing direct buys at much higher prices.

But those same savvy early adopters who realized huge discounts by buying the same users on the same publishers over the open exchange saw the writing on the wall. They recognized that prices were rising on the best users as the competition in the auction rose – since unsurprisingly, the same users seemed to be of interest to all advertisers in the same sector.

The savviest advertisers went directly to publishers and made PMP deals to access inventory with mechanisms that gave them advantage over their competition – which is also known as privilege. By putting their PMPs in increasingly higher priority within the ad server, setting up fixed-rate, variable and hybrid-rate deals and using new tools like header bidding, the most knowledgeable buyers stayed ahead of the competition. Publishers saw that these new mechanisms drove much higher CPMs, in many cases higher than direct buys, and importantly gave them insight into why advertisers bought from them. Eventually, the very most desirable audiences on the largest and best publishers evaporated out of the open auction.

The market is tipping over on itself – with open auctions being relegated more and more to purely direct-response advertisers that are not selective about which publishers their advertising runs on. For large brands, especially those spending large budgets, which also tend to be those that care deeply about running ads on high-quality publishers, things have gotten a lot more sophisticated.

Programmatic is no longer about low-cost inventory; it is now the infrastructure for transaction where the buyer and seller are handshaking and establishing connections to the consumers that brands need to reach. Programmatic is the mechanism to bind together the new tools that empower the advertiser to take control of their audiences and apply real science to the art of advertising. Publishers now can gain insight from working through these mechanisms rather than being left in the dark.

Sophisticated publishers already know this – and are driving programmatic elements or line items in their core I/Os as part of their direct business. On the buy side, the trend is for agencies to blow up their trading desks and embed programmatic buyers into direct buying teams.

This is a clear wake-up call for publishers that are still not treating programmatic as part of their direct sales or which haven’t changed sales compensation to remove channel conflict. Same for advertisers and media agencies who are segregating their programmatic buyers from their direct buyers.

Deal design has gotten extremely sophisticated, and the trend is toward increased sophistication, not simplification. If you are driving programmatic sales at a publisher and your deals are very one-dimensional, you’re probably missing something.

If you’re buying programmatically today and haven’t analyzed the core audiences you’re reaching over the open exchange, broken out by publishers that you’re also buying directly from, you’re behind your competitors.

And if you’re a marketer, question your media partners about all of these things. You have time, but not very much.

4 Comments

  1. Gerard March 10, 2017
    In effect, were seeing “networks” appear that advertisers use. Yes! ad networks are back but the publishers are acting like their own middleman. The buyers can now group together publisher networks and create their own ecosystem of their own choosing. Tis a fun time to see the same philosophy repeat itself
    • Eric Picard March 13, 2017
      Gerard, it’s not quite the same thing, neither philosophically nor structurally. Ad networks were arbitrage mechanisms designed to extract money from the ecosystem. This is a direct relationship between buyer and seller. The seller and buyer pay only technology fees and deal with the negotiation costs.
  2. R.J. Lewis March 15, 2017
    Eric,Great piece. We definitely live in interesting times. The pace of change is such that the costs of both technology, talent, and training to keep up, may out weigh the benefits gained. Are we as an industry encouraging brands to sit on the sideline and wait for the dust to settle? Also, and most importantly, what do you see as wave six?

    Reply

    • Eric Picard March 20, 2017
      Hey R.J. This is a trend that is finally culminating after a long incubation. I don’t see it as a time for anyone to sit on the sidelines – this is the holy grail we’ve all been waiting for: Publishers are empowered to sell and build value-based relationships with buyers. Advertisers get value from their customer data investments and the ability to intelligently decide who to reach, at what frequency, and how much to pay for that exposure. Wave six? I just got Wave five out to you – let’s start there.
Dec 21 2016
Leave a comment
Ad Ecosystem, Publishers

Ad Tech Vet Eric Picard Joins Pandora As VP Of Ad Product Management

Originally Published on AdExchanger.com – Wednesday, December 21st, 2016

Pandora is increasing its bet on ad tech.

The streaming music platform will bring on Eric Picard as VP of ad product management to continue building out display and video products and lead its dive into programmatic audio.

Picard is a longtime ad tech executive. In 1997, he launched Bluestreak, one of the first companies to create the rich media formats that are standard in digital today. Since then, he’s launched numerous ad tech startups, led ad product strategy for Microsoft and, most recently, was VP of omnichannel media for MediaMath. Picard joined MediaMath via its acquisition of Rare Crowds, a programmatic platform he founded in 2012.

“I’ve been in ad tech my entire career,” he said. “I have played roles in teams across pretty much every aspect of the space.”

In audio, and at Pandora specifically, Picard sees an opportunity to “participate in such a large marketplace for an ad media type that hasn’t been fully explored yet.”

“There aren’t too many places in the market to go that are nearly as exciting as the marketplace that Pandora has built for audio, display and video ads,” he said.

In his role, Picard will lead a team of 15 to 20 product managers focused on building and optimizing ad products. He plans to grow that team during his tenure.

Pandora has been bullish on programmatic display but hasn’t yet begun selling its in-stream audio ads programmatically. Picard will likely have a big part in pioneering that in 2017.

“I’ve been deeply involved in the next generation of platforms and methodologies, what we loosely call programmatic,” he said. “You can imagine that we’re thinking a lot about a lot of those things.”

Pandora offers an opportunity to innovate in an area of ad tech that’s still nascent.

“Figuring out the future of audio is obviously the enticement,” he said.

 

Sep 30 2015
Leave a comment
Ad Ecosystem, Online Media, Predictions, Publishers

Why Do Web Pages Load So Slowly In A Broadband World?

By Eric Picard (Originally Published on AdExchanger.com – Wednesday, September 30th, 2015)

If you ask anyone, anywhere, if they like advertising, the answer will likely be a laugh and quick “no.” From a small number of people, you will get a virulent “hell no!”

But most people recognize that the content they consume is free because of advertising, and they have been willing to accept the quid pro quo of free content funded by advertising for nearly all media, for nearly all time. That’s changed over the last few years, and the easy installation of ad blockers – which frequently improve the experience of viewing web pages – has negatively impacted the ad-supported Internet.

We’re in this situation as an industry because we’ve abused our relationship with consumers. We’ve failed to design pages with the user experience optimized around the content first, with the advertising experience seamlessly incorporated into that content. That is not a call for native advertising. It’s a call to actually design the advertising and content experiences together – and to ensure that both work well and satisfy consumers’ need of great content for free.

What I mean here is that the page needs to load quickly, with the content loading first, followed by ads and then invisible code to track users. In addition to loading quickly, the page needs to be beautiful and have high utility for the user.

We all have had the horrible experience of tapping a click-bait link in social media that leads to a web page with a photo gallery of 20-plus images, each of which require as many as three clicks to move to the next image. Each click also leads to a new load of advertising. That’s the most egregious example of what is frustrating consumers today. It should be equally frustrating to advertisers and agencies – who are basically footing the bill for terrible experiences and likely getting no value from those ad impressions. 

Unfortunately, the mean load times of nearly all content pages on the Internet is not much better than these “bottom-of-the-barrel” sites, with a few notable exceptions. Once you move beyond the very best publishers, the cliff over which the consumer stumbles is pretty high. The vast majority of sites don’t load much faster than the very slowest.

Why has the web become a wasteland of user experiences, and why do web pages take almost as long to load today as they did back in the days of dial-up? Is this fixable?

A History Lesson

As I’m finding more often these days, we need to look back in order to look forward.

In 1997 when I started Bluestreak, one of the first rich media advertising technology companies, the bandwidth available to nearly all Internet users was dial-up constrained to either 56K baud (that’s bits audio) or even 14.4K baud. That’s worse than your worst mobile data connection today. Yet we were able to deliver amazing ad experiences. But that was almost 20 years ago.

Bluestreak’s technology was Java applet-based and designed to support the needs of low-bandwidth users at a time when publishers had extremely conservative file size restrictions on advertising.

Our initial load of image and code was less than 1 kilobyte (KB) of data, which would render an ad on a page with a message that read, “Loading.” A subsequent download of less than 5 KB would get an initial image onto the screen. The total subsequent load of a rich media banner would be less than 64 KB. And these were rich media ads – not static banners.

In 1998 we rolled out expanding banners, rich media applications with multiple pages and all sorts of “special effects” and various interactive behaviors. The following year, we launched some of the first video ads online. We made wonderful things happen for advertisers and consumers within the very tight constraints of bandwidth and file-size limits.

With bandwidth basically unlimited today, why do pages load so slowly when we proved almost 20 years ago that great ad experiences could be loaded on dial-up connections?

Solving The Problem

Publishers really own the bulk of this problem because slow page loads relate to how pages are coded. Software for delivering web pages must be optimized such that the site’s visual components and content load very quickly. This is not an overnight change – it may require entire web experiences to be recoded. Finding quality engineers in the publishing space who understand how to code pages properly is a challenge. But this is a critical and almost existential issue for publishers, and we’ve repeatedly seen how good user experiences drive up the value of pages.

To that point, advertisers and agencies need to hold publishers accountable for the user and advertising experiences. They should stop buying advertising from publishers that don’t solve this problem, or at the very least push hard on publishers to ensure that they design pages that load quickly and are not bandwidth hogs. This last part is particularly important for mobile – where the user’s data plan is being impacted by all the content being loaded on the page, including the advertising.

When buying ads programmatically, advertisers and agencies should use a technology provider like Trust Metrics or Integral Ad Science to determine if the advertising experience being provided is high-quality and brand-safe. The technology provider can scan a web page to determine if there is quality content and page layout, with a small number of ads and sufficient white space, or if the page is an “ad farm,” with dozens of ads.

Creative agencies need to design ads that load quickly and optimize file size. This means building teams with coding skills to build fast-loading HTML5 ads and working with rich media vendors to build optimized ad experiences.

Similarly, rich media ad companies need to embrace the idea that desktop web users need fast-loading ads – even if they are on broadband – and that rich experiences don’t require massive file sizes or bandwidth.

And agencies should vet these companies and ensure that they are following best practices. While desktop users typically have “all-you-can-eat” data consumption plans, that’s not the case for mobile. Many of the pages we visit on mobile are non-optimized desktop sites that load even even more slowly over mobile devices. If the consumer’s data plan takes the hit of all the ad content loading, it’s injury to insult.

Users are not blocking advertising because they hate advertising. They hate the horrendous experience of visiting terribly coded and designed web pages with too many and slow loading ads. If the experience of viewing the web using an ad blocker is significantly better because pages load faster and look better, this is purely a problem that publishers, creative agencies and rich media companies need to fix.

Our industry is the problem, not the consumer. So let’s fix it.

Jul 08 2015
Leave a comment
Ad Technology, Online Media, Publishers

How Microsoft Almost Won Digital Advertising

By Eric Picard (Originally published on AdExchanger.com, Wednesday, July 8th, 2015)

The announcement last week that Microsoft is effectively selling off its display advertising business to AOL made me a bit nostalgic. I was recruited by Microsoft as it geared up for a major foray into the advertising space.

Although I only worked there from 2004 to 2010, I think my perspective on the company’s evolution and decision to leave the display advertising business holds some value.

When I joined Microsoft, there were 20 people on the product planning team responsible for advertising technology products. The engineering team for ads was about 400. By the time I left in late 2010, the business team had grown to more than 300, and the engineering team had more than 1,500 heads. And that doesn’t include the sales and marketing organizations.

While I was at the company, we acquired seven ad tech companies, reviewed hundreds and engaged on about a dozen. We invented whole swaths of technology that the market, in general, isn’t aware of. We drove massive innovation and investment in the space. We could have won it all.

Moving To Microsoft

I had started one of the early ad tech companies – Bluestreak – in 1997. We had raised a large war chest of venture funding – and acquired several companies after the dot-com bubble burst in 2001. In late 2004 I was recruited to Microsoft by Mike Hurt and Joe Doran.

During my interview, Joe disclosed that Microsoft had come to the realization that digital advertising was critical to its future. He showed me printed slides showcasing Google’s revenue growth, funded completely by ads. Google would soon make more money than Microsoft from each copy of Windows.

In no small part, this revelation drove the decision to fund Microsoft’s search product, especially the advertising engine behind it, referred to as Project Moonshot at the time, later to be called adCenter. AdCenter was about a year from launch, the center of innovation and scale for the company. Microsoft’s broad analysis showed that digital advertising was critical to the ongoing funding of software, which was increasingly being bonded to the Internet. Joe needed someone who understood the ecosystem and could help drive the future strategy of the company. He laid out an enticing opportunity: I could help drive the investments that Microsoft would make across the ad technology landscape.

Joe described a scenario where digital advertising was potentially a core monetization mechanism for Microsoft software products that would either serve as their primary revenue source, enhance revenue, or offset lost revenue from piracy.

Over the course of the next few years I met an extremely impressive cast of characters.* They ranged from the core business team under Joe to some of the most brilliant engineers I’ve met and executives from whom I learned an immense amount about business and technology.

Microsoft’s Not-So-Secret Weapon: Engineers

When I talk to people about the value of world-class engineers, they often fundamentally misunderstand what I’m talking about – because they’ve never worked with world-class engineers.

There’s a whole set of assumptions that are wrong, such as the belief that engineers build what business people ask them to build. Or that engineers are socially goofy and can’t understand business issues. That engineers would never get anywhere without business people who translate the market to them.

The engineers who I worked with at Microsoft – especially at senior levels – were in many cases geniuses. While there was the occasional social stumble, this was less common than you’d expect. And any of the senior engineering leaders could easily transition to CEO or non-engineering leadership roles at most companies – and many have.

In the first few weeks at Microsoft, I met a handful of engineers with whom I’d form long and fruitful relationships. Tarek Najm was the engineering leader who started the adCenter team. He’s one of the most brilliant people I’ve met – extremely inventive, high-energy and curious. Tarek took the lead in trying to catch Google’s AdWords product. With a relatively small team, he built a superior monetization engine from scratch.

One of Tarek’s lieutenants on adCenter was a program manager named Brian Burdick, who became one of the great unsung heroes of the advertising technology space. Brian is the one who ultimately invented RTB.

Tarek’s lead engineer for display advertising was a wiry man named Phani Vaddadi – who brought with him his two lieutenants, Alam Ali and Brian Tschumper. These three guys formed a back-room brainstorming group with me. Among other things, the four of us came up with some ideas around ad-funded software that we incubated and brought to market, which ultimately became the mechanism by which ads were delivered into Xbox.

There were also numerous trials in a variety of devices and applications, from the ill-fated Zune to trials of ad-funded Office and Windows in various markets across the world where piracy was an epidemic.

During my first year, we launched new brands, including Windows Live – if you can remember that one – and innovated on advertising formats. I crafted a set of principles regarding when and what kind of advertising was appropriate for which content experiences. It was based on the idea that modality of the user experience should drive whether we showed ads at all, such as when a Hotmail user is composing an email, or whether the ad could be disruptive, such as covering the page where a user is reading an email.

2005-2006: The Plan And Beginnings Of Execution

In addition to being responsible for overall ad technology strategy, I led a group focused on “emerging media.” This included mobile, OTT and addressable TV, video game advertising, device-based advertising, ad-funded software and a category known as “other.” Working with Joe, his direct reports and some of their direct reports, we crafted a comprehensive vision and plan for winning the ad technology space.

The strategy that evolved was pretty comprehensive and clear: build, buy or partner analysis on all opportunities in the space. Where we had existing investment in heads and technology, we’d increase our investment in alignment to revenue opportunity. We would acquire other companies in the space that owned strategically valuable components and held significant market share. We’d partner when there were assets that were not strategically important to own – but were needed for our customers or to operate our business.

The overarching vision was to be the platform of record for buyers and sellers, and use the scale of our technology investments to drive prices down while claiming a small percentage of all transactions. Our vision was that we’d automate buying and selling, and build direct connections between buyers and inventory owners wherever possible.

In 2005, Joe asked me to pick up all the M&A coordination work. Over the next few years, we reviewed hundreds of deals and pursued about a dozen.

Video

I engaged on a massive video and television advertising project that went through various iterations for nearly three years. Steve Ballmer had asked Joe to rationalize all the video advertising projects across the company and ensure that we had one cohesive strategy. Within three weeks I found six major initiatives across three divisions of the company that all were trying to build a comprehensive video or television advertising product suite as a standalone. It took several quarters, but eventually we rationalized all these projects and packaged them up.

I suggested that we should either partner or create a joint venture with broadcasters, networks and studios to offer a digital version of their content over the web. It would be integrated into all of Microsoft’s consumer-facing video consumption assets, including Xbox, Windows MediaCenter, Microsoft TV, Windows MediaCenter and MSN Video. This was before YouTube, while Netflix was still mailing DVDs. Our various business discussions with broadcasters may well have been the kernel of the idea for Hulu.

We had significant investment across numerous divisions and technologies – and we supported video advertising for one of the largest digital video providers, MSN Video. We invested in software to run video ads in any Microsoft product or device.

In-Game

On the video game front, Kevin Browne reached out to us while investigating the emerging area of “in-game” advertising. He said that some new companies were driving significant revenue to game studios by dynamically inserting ads into the video game, usually in a billboard-like model.

He suggested that the Xbox division wanted the capability to support in-game advertising but it wanted the overall monetization and advertising sales to be centralized outside of its team. Joe and I had agreed upon a strategic framework for technology investment such that if any player in an emerging market had gained significant market share that seemed sustainable, we should consider them for acquisition.

Massive fit that bill exactly: It had about 80% market share and was growing. While there were other companies in the space, Massive was the standout – nearly defining the category. It became the first of several acquisitions I was involved with for the company. It also taught me for the first time exactly how hard it was to get acquisitions at Microsoft to work post acquisition.

Mobile

Microsoft has obviously lost many opportunities in mobile, not least of which is in mobile advertising. But in the days before the iPhone, when the smartphone market was made up of Blackberry, Microsoft and “other,” Windows Mobile had a chance to be big.

And we saw mobile as a big part of our strategic footprint. We invested in core assets in the mobile space. To bolster our European footprint, we acquired ScreenTonic in France.

Nobody imagined Facebook back then. Nobody imagined that Apple would build a smartphone. And Google was a threat we all feared. In 2005, Google acquired Android – but nobody got it.

Programmatic

In 2005, I first heard about a paper written by Brian Burdick, with help from others on the adCenter team. He proposed something called an Open Listings Exchange (OLX) to mirror the financial markets when ad exchanges went digital. His paper was a revelation. I believe it was the first time anyone proposed the concepts we now know of as real-time bidding (RTB) to the market.

In my purview of emerging media was that category called “other.” It was in this “other” category where the OLX lived. Today, we call it “programmatic.”

The adCenter team proposed building a broad overarching platform that was open and available for all parties in the space to develop against and plug supply and demand into. When we pitched this to Bill Gates and asked for 1,000 engineers to run after this opportunity, he balked.

This led ultimately to our acquisition of AdECN, which had an early ad exchange that didn’t quite meet the technical need we envisioned for OLX. But that wasn’t until 2007.

Search

Also in 2005, Microsoft brought in David Jakubowski to build a new product marketing team for adCenter to effectively bring adCenter and paid search ads for our search engine to market. David hired a stellar team of leaders that included Brian Boland, James Colborn, Jennifer Kattula and many others. With great product managers like Jed Nahum, Erynn Petersen and Saleel Sathe on Joe Doran’s team, along with others working with David’s team, adCenter and related products and technologies went live.

What Went Wrong

Over the next few years, we significantly grew our investment in advertising technology, with much of the investment going toward our defined build, buy or partner strategy. We acquired DeepMetrix as a web analytics provider, Massive for in-game advertising, Screen Tonic for mobile advertising and AdECN as an advertising exchange.

All of these acquisitions were done with the expectation that we would bite off a big chunk of a market and grow – but as I learned, Microsoft had a hard time ingesting acquisitions at the time. There are many reasons why. Suffice it to say that DeepMetrix, ScreenTonic and Massive didn’t provide the catalysts we’d hoped for to jumpstart these marketplaces. Of all of them, only the AdECN acquisition seemed to have real promise because Brian Burdick took over engineering as CTO and ran after RTB.

2007: aQuantive

Numerous times in our strategic analysis of the space, our team recommended running after DoubleClick. Ultimately, our executive chain was unwilling to consider such a large acquisition in the 2005-2006 timeframe, so we went after other opportunities.

By late 2006, we had been pushing our vision externally to target opportunities with video and even OLX. We’d met with every large media company and every large company in the TV and video content space. Mostly these strategic discussions were driven by Yusuf Mehdi, Joe Doran, me, folks in the corporate strategy group and Tarek Najm.

In 2007 Yusuf, who had been the CVP who managed search, MSN and advertising, was promoted to the title of SVP and chief advertising strategist. This signaled internally and externally that Microsoft was very serious about investing in digital advertising. Since my team owned ad tech strategy, I was asked to dotted-line report to Yusuf as we started considering big strategic opportunities.

By 2007, with Yusuf’s promotion, we started reviewing much larger and more strategic deals and investments. We recirculated across the video content space and held numerous meetings about our OLX vision and the desire to invest in an alternative to Google, which resonated with strategic partners. Executives from agency holding companies and media companies frequently expressed extreme interest in Microsoft developing as the alternative to Google in paid search and across all digital media.

We began to get very serious about a few big acquisitions that we’d developed an appetite for. One was DoubleClick – the other was Donovan Data Systems.

DoubleClick was the only company that met our strategic framework on the ad platform side. It had a huge position – approximately 65% on publishers and about 45% on agency desktops with DoubleClick for Publishers (DFP) and DoubleClick for Advertisers (DFA). Importantly, we started hearing about a new large project internally called the DoubleClick Exchange.

We investigated and ultimately passed on acquiring Right Media at the end of 2006. We were now fervent in our belief in the OLX vision, which had matured over two years. OLX could be catalyzed by combining the supply from DFP with the demand from DFA, with Microsoft inventory as an anchor tenant. We’d have the opportunity to really take off.

We saw Donovan Data Systems as a perfect fit in our strategy. It had a huge percentage of agency media buyers using its systems, and was a big Microsoft customer.

Unfortunately as we neared a swing at DoubleClick, which would have been the centerpiece of our strategy, it ran a quick process and stepped into exclusivity with Google. We tried unsuccessfully to break them out of that exclusivity and were prepared to throw a ton of money at it – but Google prevailed.

The alternative approach that Yusuf, corporate strategy, Joe and I came up with was less than optimal. We’d basically acquire and roll up several major assets. We bought AdECN to create a center of gravity around our OLX vision. We continued discussions with Donovan Data Systems and got very close to a deal.

And we began conversations with aQuantive.

Since aQuantive was based in Seattle, it was easy for our executive team, who hadn’t been deeply ingrained in the strategic view so far, to step in and participate directly in conversations. And things accelerated quickly – so fast that negotiations moved beyond the pale of expectations – with the valuation of aQuantive eclipsing the next most expensive acquisition at Microsoft by a wide margin.

Ultimately, Microsoft decided that aQuantive was the big bet we would make in the space. The strategy was to leverage the buy-side footprint of Atlas, which was similar to DoubleClick’s 45% market share, and attach it to the AdECN exchange to form the basis of OLX. While I continued to push hard for Donovan Data Systems to augment that Atlas footprint, the decision was made to focus on aQuantive and build out an automated optimization engine that would connect Atlas with AdECN, providing automated bidding capabilities. Microsoft’s ad network inventory would anchor the exchange, including owned and operated remnant inventory with a small amount of premium inventory. And we would create synergy with our existing adCenter customers.

Things didn’t proceed as planned. It took a long time to get the new aQuantive team up to speed on our OLX vision, and they were skeptical. The aQuantive leadership team became the business leadership team of the new advertising organization that swelled to 1,500 engineers and 300 business people The aQuantive executive team never embraced our OLX-enabled advertising platform business strategy – they felt that the astronomical price we paid for the company validated their previous strategic direction. They felt strongly that we needed to incrementally grow revenue from our base, which is how they’d grown their company. What they missed was that their existing revenue had very little impact on the strategic imperatives that Microsoft cared about. We needed to move the needle by billions of dollars, not millions.

The plan had been for Yusuf to lead the new division, with his core leadership team making up the leadership ranks. During the final stages of the aQuantive negotiations, a new path was forged with Brian McAndrews and his team stepping into the lead. I really liked those guys – and had been friends with many of them for years ahead of the acquisition. But ingesting and digesting that acquisition was really hard for both companies. And adECN died on the vine of that ingestion. We weren’t allowed to start testing live inventory through the exchange because an executive wouldn’t sign off on the revenue risk.

Ultimately we lost our opportunity. Prior to that acquisition, we refuted the idea that Microsoft couldn’t be agile and responsive to the market. After the acquisition, we crawled into our cave to digest a big meal – like a dragon. By the time we emerged from our cave, the world had evolved past us.

We ran instead after a giant partnership with Yahoo on search. We reduced our investment in display and other forms of advertising. That defocus culminated finally in the exit we saw last week from everything but search and paid search.

But there was a time when Microsoft almost won. We were duking it out with Google and focused on a major win, not just participating. We led the market. Many of those in this story went on to huge careers in advertising – with several now at Facebook.

We almost had it.

* While posting a comprehensive list of people on Joe’s team back then would be nigh impossible, there are some key players that should be mentioned. Those included Alexandra Tibbets, Jed Nahum, Michael Dwan Matt Carr and Mike Hurt and Some real powerhouses that worked under them, giving the bench on this team extraordinary quality and depth, including Ryan Mackle, John Genna, Meera Bhatia, Sloan Ginn, Aaron Sandorffy, Michael Weaver, Dean Carignan, Gabriel Nanda, Gabe Bevilacqua, Mark Jacobson, Gary Hebert, Jilani Zeribi, Khan Smith, Erynn Petersen, Saleel Sathe, Maziar Sattari, Jenn Dorre, Bart Barden and Matt Romney, as well as many others I’m sure that I’m forgetting, with apologies.

14 Comments

  1. Augustus July 8, 2015
    Are these the same “world-class engineers” that wanted to convert DRIVE to run on Microsoft’s in-house ad server that couldn’t support flash, CPC/CPA cost methods, or 3rd party publisher inventory in 2007? Or the ones that claimed adCenter was fully “converged” and display capable in 2008? Or maybe it was the ones who attempted and failed to build a publisher ad serving system from scratch after spending 6 billion to acquire a company that had all these pieces. Let’s not shit ourselves, the failures were abundant. Trying to pass it off as aQuantive leadership’s inability to see a larger, Microsoft-wide vision is to ignore the inherent flaws in the Microsoft strategy you claim to have helped craft. Did you really think a network or exchange anchored with 90%+ Microsoft owned and operated inventory was going to be a solid platform play? Were you seriously banking on converting demand from adCenter to spend in display? Was keeping Microsoft’s targeting data confined to O&O inventory and off the network (and ultimately the exchange) just something that was done because everyone got in a room and decided that they hated making money?You’re right that we almost had it. If it weren’t for that $6 billion, we (AQNT) would still be having it.
    • Eric Picard July 9, 2015
      Hi Augustus, nice to hear from you. Let me avoid a back and forth snipe-fest and just address a few factual issues with what you said, and maybe answer a few of your questions.1. “couldn’t support flash, CPC/CPA cost methods, or 3rd party publisher inventory in 2007?”That’s actually wrong on all counts with one caveat. Microsoft’s Display Ad Platform is (and was) a pretty remarkable platform. It was not designed to support external users logging in – which was literally a user permissioning and data segregation issue. Frankly – that’s not a problem that required world class engineers to solve.

      2. “Or the ones that claimed adCenter was fully “converged” and display capable in 2008″

      adCenter was never claimed to be “converged”. I don’t recall the date we began supporting display ads in adCenter (actually the pubcenter product) but I don’t believe it was 2008. Given that the convergence project (systems integration) was literally never completed, and there’s plenty of reasons I could give for that (e.g. plenty of blame to go around), that’s just a silly statement. Many core systems became shared, but obviously since Atlas was able to be sold off, it remained standalone.

      3. “inherent flaws in the Microsoft strategy you claim to have helped craft. Did you really think a network or exchange anchored with 90%+ Microsoft owned and operated inventory was going to be a solid platform play?”

      The market clearly has shown that companies *without* the vast volume of inventory Microsoft could have passed into an exchange were able to be very successful both before and after the timeframe I’m talking about (Right Media and AppNexus are obvious examples) your point doesn’t make much sense. AppNexus really took off after he additional supply from Microsoft was added. So yes – I think it was a very solid platform play. The market shows that to be true. Obviously Google made lots of rain with the DoubleClick platform as well – but given that there are other examples (Rubicon, Casale, OpenX, AppNexus) yes, Microsoft certainly could have done it. Given that we had solicitations from dozens of huge publishers and literally every major agency holding company, who literally asked us to build such a platform, yes – I think we could have done this.

      4. “Were you seriously banking on converting demand from adCenter to spend in display?”

      Banking on it? No. But was it applicable? Obviously it was – Google was clearly able to apply its AdWords demand against display (e.g. Google Display Network.)

      5. “Was keeping Microsoft’s targeting data confined to O&O inventory and off the network (and ultimately the exchange) just something that was done because everyone got in a room and decided that they hated making money?”

      I’m not going to name any names. But this was literally the plan of record prior to the aQuantive acquisition. The plan of record was to open up all MSFT targeting data (effectively offer a DMP) and all inventory short of a set of premium established inventory onto the exchange. So you’d need to tell me the answer to your question.

      What I was told at the time was that doing so would put too much revenue risk on the O&O inventory to even allow a few million of impressions come out of hotmail to run adECN tests. And there was a lot of discussion about liquidity and asymmetry that would have been easily addressed if we were allowed to actually run tests. Not sure what more can be said about that.

      Eric

      • Augustus July 10, 2015
        Eric, we’re talking about a scenario where thousands of people lost their jobs or at least had their careers significantly derailed as a direct result of terrible strategy + execution. So I agree, let’s get the facts straight.1. “couldn’t support flash, CPC/CPA cost methods, or 3rd party publisher inventory in 2007?”This is just a fact. Even up until about a year ago, AdExpert couldn’t support performance cost methods. I’m not even sure it can today. Point is, shortly after the acquisition, there was an engineering-led effort to convert DRIVE (a top 5 ad network at the time, mind you) to AdExpert by Microsoft. I personally was asked directly by the MSFT engineering team which of those features (among a laundry list of others) the network could live without, preferably all 3, was how it was phrased.

        2. “Or the ones that claimed adCenter was fully “converged” and display capable in 2008″

        Again, this happened. I won’t name names either, but let’s just say the head of engineering at the time announced exactly this statement at an all hands. I was there. A few months later, he left the company and we all discovered that this claim was without merit. A silly statement, I agree.

        3. Let’s not act like Right Media is a shining example of platform success, and I think AppNexus would be just fine without Microsoft inventory. Look, the anchor tenant plan was a great one, I’m not arguing that. But when the anchor tenant is the only tenant, you don’t have a platform. The publisher tools business was established and growing within aQuantive (and RAPT), and shortly after those acquisitions, the strategy to move all of those pub-side tools to a different platform is what killed it. Publisher customers were FIRED, if you recall. Fired them. They were paying money, Microsoft said, “nah, don’t want that business.”

        4. Search and display were separately managed budgets then, and they still are today. Again, just a fact. If the plan was to change the way the industry spends across these 2 formats, you needed a lot more than one more checkbox in the adCenter UI.

        5. Great, something we can agree on. Tell me this then, why was targeting data kept off of DRIVE immediately after the acquisition, and remained off of the AdMarket platform for the remainder of its existence? I can send you a Quick Wins doc where this is laid out clearly as something that would have made an immediate revenue impact within 90 days, and yet it was promptly shut down by Microsoft leadership… on the engineering side, btw.

        Your vision for adECN at the time was indeed a great one. Missteps were made by several folks (I know the ones you are referring to) that prevented the exchange strategy from taking hold and flourishing. But Google has been able to successfully execute a display network and an exchange, both best in their respective classes. That combined vision is something that was absent throughout the process, or at least never agreed upon in a way that allowed for successful execution. Those of us in the rank and file felt most of the pain resulting from these decisions and lack of solid leadership. It would be nice if ALL of the leaders responsible took their fair share of accountability for the disaster.

      • Eric Picard July 13, 2015
        Augustus, as I feared things went down a didactic path. So let me try to address the intent of the article rather than going back in and picking apart your reply to my reply.My motivation for writing this article was that the press response to the AOL announcement was to basically repeatedly state, “Yeah, Microsoft never knew what they were doing in advertising.” That simply is not true. The strategic blunder that the company made was in acquiring aQuantive and losing three years that they were never able to recover from. This was what I was alluding to in my article by referencing the digestion of a meal that was too large. Keep that in mind when reading my further comments below.Since I left Microsoft in 2010 (when it became clear to me that the company was not going to continue to invest in anything ad related but Paid Search) I no longer have access to any of the direct paperwork such as various presentations, and internal memos – many of which I wrote. But having a semantic argument about what was said by who at a meeting in 2008 at this point seems superfluous. I’ll just say that I wrote most of the decks that were presented at engineering leadership / all-hands meetings post-acquisition – and I think your memory and mine are very different.

        I will address one thing that I haven’t, which you brought up in both of your comments – regarding the publisher tools business. aQuantive had acquired Accipiter and renamed it Atlas for Publishers (or something like that.) I have nothing but respect for Brian Handly and the many folks from Accipiter that I knew over the years. But that platform was ancient and architecturally needed a complete rewrite. It was simply not possible for that platform to be the center of gravity of the business going forward. You reference this as if it was as simple as attaching RAPT to Accipiter and backfilling with DRIVE PM. That wasn’t going to work on any level – just the handful of sales done with large publishers after the acquisition proved that Accipiter wasn’t salvageable. It’s unlikely you were aware of those issues, but I can tell you without any hesitance that this wasn’t going to work.

        I can also tell you definitively that the decision to move away from plan of record on the post-acquisition timeline was not made by engineering. I was in those meetings. Your perspective is missing key facts – but I’m not going further on that.

        My point in this article was not to point fingers at anyone and blame them for Microsoft’s subsequent failure in “non-search” advertising. I have huge respect for Brian McAndrews, Mike Galgon, Karl Siebrecht and Scott Howe – they’re all very talented and intelligent executives. If this article seemed like it was taking pot shots at them – that certainly wasn’t the intent. See my comment above about dragons and meals.

        The issue simply is that there were vast and complex systems across both companies, and a consensus based decision about which systems to bet on was allowed/caused to go on for more than 2 years. The big lesson to bring away (although I was Cassandra in this one – having learned this lesson earlier in my career) is that clear definitive executive decisions about paths forward (whether engineering or business) need to be made quickly and followed through on. But that wasn’t the point of my article – so perhaps there’s another article in there about how to do acquisitions well and what to avoid.

        I will respond directly to one statement you made, “But Google has been able to successfully execute a display network and an exchange, both best in their respective classes. That combined vision is something that was absent throughout the process, or at least never agreed upon in a way that allowed for successful execution.”

        This is exactly the point of my article – the entirety in two sentences. Our vision of the future of the market was exactly the same as Google’s prior to the aQuantive acquisition. And that vision was shared across engineering and business from the lowest to highest levels. Unfortunately it took more than two years to get that vision accepted and understood across the executive team post-acquisition. And that’s the tragedy of Microsoft’s advertising business, the lost years while the market surged past us. When it was clear that nobody was going to bless adECN as the exchange for Microsoft, I didn’t raise any objections when Microsoft bet on AppNexus. At least there would be one platform in the market that matched our overall objectives – and we’d own some of it.

  2. Robin Laylin July 8, 2015
    Eric, thank you for taking the time to describe this period at Microsoft, one with so much potential for not only advertising business pursuits, but also benefitting and leveraging Microsoft Enterprise identity, server, desktop, analytics products to deepen reach and value. Thanks again for the excellent summary!
  3. Alan July 8, 2015
    Great write up as usual Eric. This tale reminds me of a Yankee fan talking about how great their team would have, could have been if only this that or the other had happened. And since they bought the superstars, had all the resources in the world, unlimited funds and still sucked whose fault is that?I am amazed at how much money Microsoft threw at this industry and lost. To me its a lesson in how not to run a business and a very real example of how large companies are seldom, if ever able to compete in emerging businesses.
    • Eric Picard July 10, 2015
      Alan – thanks for your comment! I absolutely hear you. But the reality is that this is more along the lines of Xerox Parc lamenting the Graphical User Interface being credited to Apple. 😉My main motivation for writing this was that most press I read in response to the AOL deal got this all very wrong. Repeatedly I was reading sentiment stating that “Microsoft never knew what they were doing in advertising.” That’s just simply not true.Microsoft’s in-house team was ahead of the market curve. And we were executing well toward that plan. Not without missteps, mind you. But there seems to be a sentiment that Microsoft didn’t know what to do in the ads space, which isn’t true – we were doing very well.

      As far as the lessons you suggested – it’s really hard for big companies to take on new challenges and succeed. So we’re in agreement. But Microsoft has built more large new businesses than any other company – so it can be done – the question is how to do it.

      Microsoft’s past had shown that big moves with either huge internal investments with giant teams (Office taking over the world or the huge investments and losses of Xbox before it became profitable) or large acquisitions driving big new incremental businesses (Great Plains driving MSFT’s enterprise business forward) were good patterns. But Microsoft had never faced a competitor like Google before – and they proved impossible to fast follow against. The Bing investment turned out to be much more like Xbox than Office.

      • Alan July 10, 2015
        Thanks Eric for your response. The issue is that neither the video game console market or search were emerging markets. They were both well established businesses for many years with many parties at play. So while I appreciate your response, it doesnt actually hold water.MSFT screwed up with a massive amount of enterprise level failure. My dealing with the company during this time (and there were many from several different companies) was one of arrogance and hubris. You guys thought you were smarter than everyone else (not you mind you, you were and still are very kind and humble guy). But that kind of arrogance always translates into failure. And boy did it ever in this regard.I realize that you gave MSFT a lot but they didnt give you what you truly needed. The reigns. And that is one of many reasons that they lost big time. But most of all is that they had no business entering into the ad business. I think that was and will always be their failure. Trying to muscle their way into a sector that really was about as far removed from their core competency as possible.

        On another note: I wish AdExchanger had more dialog on their site. This discussion is one of the best I’ve read here but they dont promote dialog between the writers and the readers and certainly dont provide a channel for engagement.

      • Eric Picard July 13, 2015
        Alan – thanks for your thoughtful reply to my reply. 😉I don’t think Display advertising was much of an emerging market at that time, but of course the movement toward exchanges and real-time bidding was an emerging space.I’m not sure who you dealt with at Microsoft in those days, but I will tell you that I was repeatedly surprised at the lack of arrogance and hubris I experienced across the board while working at Microsoft. Not to say there weren’t egos – but your experience was not the one I had.

        Microsoft was on a great path from 2004 – 2007 and making great strides toward an epic head-to-head battle with Google. But the lost years that happened after the aQuantive acquisition were not possible to recover from.

        Note – I firmly believe that if aQuantive hadn’t been acquired, they’d still be a successful business in the adtech space. So the tragedy cuts in both directions.

  4. Pat Dunbar @patdunbar July 8, 2015
    Robin, who was another unsung hero in this saga is definitely right in his congrats for Eric on the story. There were so many other non pursued threads – around the world was part of the shame of it.
  5. Realist July 9, 2015
    Eric, it took some courage to write what you did. Don’t let the haters get you down. I agree more with Augustus, most the genius of Microsoft engineers is in building three-legged stools, re-inventing wheels and blowing through budgets. Moonshot projects that need 1000 engineers? Sorry but ad tech ain’t NASA.
    1. Eric Picard July 9, 2015
      Hi Realist. The reality is that large scale teams at huge companies frequently are less efficient than smaller companies. And remember at the time we were competing with Google, who had well more than 1,000 engineers working on the project. Microsoft’s “fast follower” approach that had worked well for all its major successes previously (e.g. Office) set the stage for large resource requests.Obviously we didn’t throw 1,000 engineers at adECN when we completed that acquisition. And we did bring the first RTB exchange to the world – unfortunately we just were not able to get it launched. It sat fallow for two years before it died on the vine.Again – obviously you never spent any time with the kinds of folks I’m referencing. If you’d spent any significant 1:1 time with Tarek Najm, Brian Tschumper, Sachin Dhawan, Nitin Chandel, Subir Sidhu, Scott Tomlin, John Beaver, etc… you wouldn’t feel like you do. My guess is you didn’t spend any time with the core engineering team at Microsoft. Your perspective isn’t informed by reality.

Jun 03 2015
Leave a comment
Ad Ecosystem, Ad Technology, Economics, Predictions

Why Programmatic Budgets Will See Massive Growth

By Eric Picard (Originally Published on AdExchanger.com – Wednesday, June 3rd, 2015)

There was a time when advertising was a game of statistical assumptions about the types of people who were consuming media. Television had four networks and there were only dozens of mainstream magazines, typically one local newspaper read by a large percentage of adults and various radio stations in each market.

In what is possibly the most basic truth of the media industry, the fragmentation trend has continued with a constantly growing number of media vehicles against which smaller slices of people’s time are applied.

Even when media-buying teams were specialized by media type, such as TV buyers and magazine buyers, the fragmentation problem still faced an unmanageable outcome. But digital media has blurred the lines between channels. Digital media buyers are now responsible for buying display ads on PC web, mobile web, digital video on both and, increasingly, audio ads. Channels, like in-game ads, and format variances, such as native ads, increase the complexity.

Billions of dollars have been invested in the next generation of media-buying technology over the past 10 years. As expected by those investors, the digital media space has grown incredibly.

The amount of money spent on digital on PCs has almost caught up to the amount of time spent by people consuming digital media – which means that spending “growth” is slowing on a year-over-year percentage basis. But spending is still growing at incredible rates. Mobile still has a massive growth opportunity that looks much like the “Internet” looked 10 years ago, as you can see below in Mary Meeker’s most recently updated “% of Time Spent” chart.

ericpicardchart

The New Planning And Buying

When planning and buying was tied to a small number of media channels and publishers per channel, it was reasonable for planning and buying group of 100 people to execute large budgets against a relatively small number of publishers. With fragmentation, the complexity of executing in any one channel makes this approach untenable.

And yet, the vast majority of ad dollars spent today are still spent against media that is bought the same way it was 10 years ago. Meanwhile, programmatic media-buying platforms have exploded on the scene and made it possible for one buyer to effectively input buying rules that allow for hundreds of billions of buying decisions per day. Each impression is evaluated in real time, valued against the campaign goals and only purchased if the value of the impression is higher than its price. This revolution puts the advertiser/buyer in control of defining, evaluating and valuing the ad inventory – a highly desirable transition to advertisers.

Although this is a technological miracle, these programmatic buying platforms have been relegated to only a small percentage of overall digital media budgets. Yes, programmatic is a rapidly growing percentage, but still has been largely limited to direct-response budgets until relatively recently.

It makes sense that direct-response budgets are directed toward the programmatic channel – buying platforms can evaluate audiences and apply explicitly identified audiences to a specific set of criteria, measured against explicit ROI goals. For direct-response campaigns, it’s easy to justify spending more than 20% of the media budget on programmatic because first-party data is such an obvious leap for marketers.

However, we have these amazing platforms with immense capabilities for evaluating enormous numbers of impressions per second and making intelligent decisions about which impressions to buy. And we have all sorts of bridging technologies and measurement models, such as Nielsen’s OCR and comScore’s VCE, to help drag budgets that need to move evolutionarily from the panel-based model approach to TV buying to more automated buying models. But there’s a chicken-and-egg problem that hasn’t been resolved.

While programmatic buying platforms are orders of magnitude more advanced than the old ways of buying media, planning methodologies for allocating budget ahead of the buying process simply haven’t kept up with the buying revolution. Under the current model, planners divvy up the budget to different buying teams, sending large chunks to “traditional” digital media buyers (an oxymoron if ever there was one) and smaller chunks to the programmatic buying team.

This is despite the fact that programmatic buying methodologies can execute both budgets equally efficiently and effectively. Buyers can just as effectively execute their budget for direct buys programmatically. The difference when a programmatic buying platform is used is that every impression can be evaluated against the campaign goals expressed by the planner, and either be bought or rejected. This “outcomes-based” buying actually puts the planner’s objectives right at the center of the buy – and pushes the media toward an even playing field between brand and direct response.

To execute a media plan using only direct buys today means that the old-world scale issues apply: A media-buying team of 100 people typically buys from between 50 to 60 publishers. This ratio means that in a world with millions of websites, a tiny fraction of available inventory is considered. And buying teams that only buy direct are unlikely to evaluate publishers outside of their personal experience, as is human nature. This is not to say that there is no place for direct media buys – they absolutely serve a purpose. But there are many other ways to run after any campaign objective, whether the desired outcome or goal is to drive an immediate sale or to reach a specific audience, or to reach a more general audience.

The Role Of Direct Buys In A Programmatic World

Programmatic buying teams now use mechanisms like private marketplace deals to execute direct buys with publishers, which enables buyers to establish more controls over how impressions will be selected or rejected than a direct buy. In a standard direct buy, every impression must be consumed. In a programmatic-first world, only impressions that match the campaign goals are purchased. And the role of a direct buy has more to do with ensuring that an advertiser can purchase inventory from a specific publisher that may otherwise be unavailable or in short supply over the open RTB inventory channel.

In a programmatic-first world, campaigns are begun over just open RTB. Using white lists and evaluating which publishers saw impression volume periodically can show how much inventory is available on that publisher over the exchanges. A private marketplace should be considered if a publisher is determined to be valuable and inventory volumes do not respond to increasing the bids on a CPM basis for available impressions. One way that programmatic-first buyers will make evaluations regarding private marketplace buys or even direct buys is to test on the exchange first to see if the inventory can be bought there. If standard bids aren’t finding the inventory desired on a publisher, and raising the bids doesn’t open up inventory, a private marketplace buy or direct buy is the answer. But there is a lot of value in finding that inventory on the exchange if possible.

It is sometimes the case that various business rules will render a publisher or set of desirable inventory inaccessible to a specific advertiser over the exchanges. In those cases, the issue isn’t bid price – the inventory is simply not accessible to the advertiser over the exchange without a private marketplace buy in place. These private marketplace deals will eventually replace direct buys. But in some cases, publishers may simply require a direct buy because their operations teams haven’t sorted out how to support private marketplaces or for philosophical reasons.

This last scenario is quickly evaporating from the market – buyers are increasingly demanding and receiving support for private marketplace deals across most publishers. It is not unusual for these to be part of a standard IO. For those publishers that require vendor support, the options that support programmatic sales are rapidly increasing. Publisher programmatic vendors, including Pubmatic, Casale and Rubicon, offer support for standard private marketplace buys. Google, as always, is innovating like crazy in this space. And upstarts, such as Sonobi and C1 Exchange, are examples of a new type of publisher-facing programmatic vendor that supports more flexible inventory guarantees, using programmatic pipes by integrating directly into the publisher ad server.

We’re on the cusp of a massive revolution in media planning and buying – with new tools and methodologies. There are significant advertiser and publisher benefits to sorting these issues out. But this innovation comes at a cost. Evaluating hundreds of billions of daily impressions across all these platforms, publishers, advertisers, campaigns, insertion orders, line items, placements and creatives is technologically intensive.

And while automation is often touted as a way to increase efficiency, that doesn’t mean it reduces labor costs. The number of people needed to execute this way stays static, but the salary costs go up because team members have more technical skills and are in high demand. But over time they scale exponentially better than traditional media buys. This will ultimately lead to some interesting conversations between agencies and advertisers. The procurement-driven cost-plus model is not leaving agencies room to support these newer and better ways of servicing their clients.

Apr 16 2015
Leave a comment
Ad Ecosystem, Ad Technology, Online Media, Publishers

A better way for publishers to manage ad inventory

By Eric Picard (Originally Published on iMedia – April 16, 2015)

Publishers in general have, up until recently, thought of programmatic advertising only as a mechanism to clear unsold (remnant) inventory. Over the last few years, publishers have been able to begin integrating their programmatic sales more completely into their overall inventory pool. And those publishers that dived fully into the programmatic pool have been gathering significant learnings and gaining sustainable advantage over their competition. For those publishers who have not fully adopted programmatic methodology into their mainstream revenue operations, the time has come.

Today I’ll be using Google’s DoubleClick for Publishers and Ad Exchange as the examples of how publishers are operating. But other ad servers, SSPs, and exchange technologies support similar functionality to what I’ll describe here. I’m using Google’s because, frankly, its documentation is public, easily found via a search (shockingly), and easy to understand. If you’re using different vendors, feel free to reach out to them and ask about these concepts. I’m certain they’ll be able to accommodate you with similar approaches on their platforms.

Starting with the basics

RTB and direct make use of different infrastructure for decision-making, and ultimately it’s the publisher ad server that “owns” the direct ad sales, which controls the destiny of whether an ad impression is available to be purchased on the exchange.

Below is an example of how ad calls are made when a user visits a web browser and the page loads. This fundamental of our business should be understood before we dive into the deeper arcana of how programmatic systems interact with the publisher ad server.

When a user visits a web page, myriad events take place — most of which we’ll ignore in this article. The important thing to understand is that publishers code ad tags into their web pages, which call out to the publisher ad server. The publisher ad server returns unique identifiers to the page that tell the browser where to find the ads that have been selected.


This is how nearly all ads are served online today — and have been for more than 15 years. What’s important is how this is fulfilled under the surface of the impressions. There are numerous interactions happening within the publisher ad server, and the external systems — including standard ad platforms like third-party ad servers (DFA, Atlas, Sizmek, etc.), dynamic creative and rich media platforms (Flashtalking, PointRoll, etc.), and programmatic platforms such as supply-side platforms, ad exchanges, and demand-side platforms.

More advanced scenarios

All sorts of decisions are made in the milliseconds between the user visiting the publisher’s web page in a browser, and all of these various systems interact with each other. But we’ll leave most of these interwoven interactions aside for this discussion and keep to the critical ways that the publisher ad server interacts directly with whatever programmatic integration it has made.

Most of the time the publisher ad server interacts with an SSP (Rubicon, PubMatic, etc.) or directly with an ad exchange (Google’s AdX, AppNexus, etc.). While I’m giving examples in some parts of this article to illustrate the kinds of companies seen in the space, the reality is that the lines are very blurry, and some might argue that components of AdX and AppNexus operate like an SSP, and components of Rubicon and PubMatic operate like exchanges. Think of them as relatively interchangeable at this point.

Regardless of what vendor and mechanism is used for the programmatic supply integration (and often multiple are used), the publisher’s ad server interacts in somewhat specific ways with these systems. So let’s begin with the prioritization queue set up within the publisher ad server.

Most publisher ad servers provide functionality to allow the ad operations teams to assign the various contracts (insertion orders, or IOs) and specifically their subsequent contract line items against specific prioritization levels within the ad server. DFP has 16 levels of prioritization available, with the first 11 levels being set aside for “reserved” or “guaranteed” line items. Of these top 11, typically the first three levels are used for sponsorships — as the highest priority line items placed into the ad server.

← Older posts