Why The Banner Ad Is Heroic, and Adtech Is Our Greatest Artifact


Every good story needs a hero. Back when I wrote The Search, that hero was Google – the book wasn’t about Google alone, but Google’s narrative worked to drive the entire story. As Sara and I work on If/Then, we’ve discovered one unlikely hero for ours: The lowly banner ad.

Now before you head for the exits with eyes a rollin’, allow me to explain. You may recall that If/Then is being written as an archaeology of the future. We’re identifying “artifacts” extant in today’s world that, one generation from now, will effect significant and lasting change on our society. Most of our artifacts are well-known to any student of today’s digital landscape, but all are still relatively early in their adoption curve: Google’s Glass, autonomous vehicles, or 3D printers, for example. Some are a bit more obscure, but nevertheless powerful – microfluidic chips (which may help bring about DNA-level medical breakthroughs) fall into this category. Few of these artifacts touch more than a million people directly so far, but it’s our argument that they will be part of more than a billion people’s lives thirty years from now.

There is one exception. The artifact we’re investigating is already at massive scale, driving billions of dollars in revenue and touching every person whose ever used the Internet. That artifact is currently called “programmatic adtech,” and it is most famously illustrated by Terry Kawaja’s Lumascapes (and less famously, my own “Behind the Banner” visualization).

lumascapedisplayYes, this is the infrastructure that allows a pair of shoes to chase you across the web. How can it possibly be as important as, say, a technology that may cure cancer? Because I believe the very same technologies we’ve built to serve real time, data-driven advertising will soon be re-purposed across nearly every segment of our society. Programmatic adtech is the heir to the database of intentions – it’s that database turned real time and distributed far outside of search. And that’s a very, very big deal. (I just wish I had a cooler name for it than “adtech.” We’re working on it. Any ideas?!)

Think about what programmatic adtech makes possible. An individual requests a piece of content through a link or an action (like touching something on a mobile device). In milliseconds, scores of agents execute thousands of calculations based on hundreds of parameters, all looking to market-price the value of that request and deliver a personalized response. This happens millions of times * a second,* representing hundreds of millions, if not billions, of computing cycles each second. What’s most stunning about this system is that it’s tuned to each discrete individual – every single request/response loop is unique, based on the data associated with each individual.

Let me break that down:

1. A person indicates a request: a desire, an intent, a preference – The Request

2. Billions of compute cycles and sh*tons of data are engaged to process that desire – The Process

3. A personalized response is generated within 100-250 milliseconds. – The Response

At present, the end result of this vastly complicated “Request Process Response” system is, more often than not, the proffering of a banner ad. But that’s just an artifact of a far more interesting future state. Today’s adtech has within it the glimmerings of a computing architecture that will underpin our entire society. Every time you turn up your thermostat, this infrastructure will engage, determining in real time the most efficient response to your heating needs. Each time you walk into a doctor’s office, the same kind of system could be triggered to determine what information should appear on your health care provider’s screen, and on yours, and how best payment should be made (or insurance claims filed). Every retail store you visit, every automobile you drive (or are driven by), every single interaction of value in this world can and will become data that interacts with this programmatic infrastructure.

OK. Let’s step back for a second. When you think of this infrastructure, are  you concerned? Good. Because it’s imperative that we consider the choices we make as we engage with such a portentous creation. This year alone, each human on the planet will create about 600 gigabytes of information, and that number is growing rapidly. What are the architectural constraints of the infrastructure which processes that information? What values do we build into it? Can it be audited? Is it based on principles of openness, or is it driven by business rules and data-structures which favor closed platforms? Will we have to choose between an oligarchy of “RPR vendors” – Google, Facebook, Microsoft – or will we take a more distributed approach, as the original Internet did?

These questions have been raised, and continue to be well articulated, by LessigZittrainWu, and many others. But we’re entering a new, more urgent era of this conversation. Many of these authors’ works warned of a world where code will eventually augur early lock down in political and social conventions. That time is no longer in the future. It’s now. And I believe as goes adtech, so goes our social code.

“Adtech” is a very important, very large application we’ve built on top of the platform we call “the Internet.” It’s driven by the relentless desire of capitalism to turn a profit, yet (so far) it has leaned toward the Internet’s core values of openness and interconnectivity. Thanks to that,  it’s suffering some endemic maladies (fraud comes to mind). It’s still a very young, relatively immature artifact. But so far, it’s more open than not. I’m not certain that will always be the case.

My argument boils down to this: What we today call “adtech” will tomorrow become the worldwide real-time processing layer driving much of society’s transactions. That layer deserves to be named as perhaps the most important artifact extant today.

Given adtech’s rise, let’s not forget its atomic unit of value: the oft-derided banner ad. In time the banner as we know it will most likely fade away, but its place in history is certain. One generation from now, we may not “click” on banner ads, but we’ll always be pulling into traffic, filing health insurance claims, buying clothes in retail stores, and turning up our thermostats. And those myriad transactions will be lit with data and processed by a real time infrastructure initially built to execute one pedestrian task: serve a simple banner ad.

55 thoughts on “Why The Banner Ad Is Heroic, and Adtech Is Our Greatest Artifact”

  1. Curious to see where you go with this. I have always said that what we do in ad tech will eventually eat the whole media world, but you are clearly taking it to the next level.

    I think the key word you are looking for is personalization. We come from a time when everyone watched the same tv show at the same scheduled time and every got the same ford f150 ad. Now each individual consumes content selecting from an infinite pool and we respond with a personalized ad. In that world you need less ads as they work orders of magnitude better. Ad tech enables data leverage to achieve utility through relevance

      1. John – Great post. What % of pubs today take advantage of data to program more relevant content and a more personalized consumer experience? Your guess.

      2. I’d guess it’s very low. Because it’s not easy to get those kinds of meaningful consumer insights as a smaller publisher. That will change in 2014.

  2. John

    I love your big picture, and forward facing thoughts. The infrastructure will be an amazing tool for so many applications. Its ironic that the marketing industry part of the pioneering. Its also ironic that more industries, governments, and various organizations haven’t utilized the potential.

    1. Thanks. I think it’s almost predictable that commercial intent – raw capitalism – would build this artifact.

  3. I’m curious if you see something like Amazon AWS commoditizing the tech/infrastructure behind adtech and if you think it’ll have an accelerant effect. I thought about that as I read your article… Would Google, MS, etc have a lot to lose if anyone could create their own adtech and tie it in with their own systems? Or maybe that just broadens the entire market (rising tide..).

    1. Amazon is already deep into adtech with its A9 unit, and is a major player in buying and selling across exchanges, and of course has its own adtech stack. It’s the quiet player, taking in loads of data and utilizing it’s first party assets to great advantage. I predicted early this year it would have a run rate ad business higher than Microsoft by year’s end, I have to take a look into that. I think this space is ripe for many disruptions – and Amazon would be one player who could do it.

  4. Wow! Takes an ‘open’ insider to tell it like it really is.

    It looks as if programmatic adtech consists of hundreds of companies each in a specific area of the total landscape. How do you see this playing wrt an 800lb gorilla end-to-end player turning up or more players entering or something else to service non-ad markets?

    When you say the infrastructure is more ‘open’ than not, what does this mean ie. are the interfaces to each area of the total system open?

    1. I mean that the ecosystem is open to new entrants – anyone can create value and feed into and off of the system. And yes, there are already massive “gorillas” in this space -Google is the biggest by a long shot. But I think Google needs this ecosystem to exist, for many reasons – and I very much believe the world will be a better place if no one player owns it.

      1. Going forward, would you say better personalization (ad relevance or response-relevance) will remain important? Also, in the Lumascapes sector or your Behind the Banner presentation, where does the ad relevance work take place?

  5. really enjoyed this John. Ad tech and the more recent programmatic feels to me more like a use case of open platforms, low cost infrastructure and application of data. The pull of open vs closed continues to play out. The interesting question is what role will the government play proactively or reactively. Have you thought about what the role of the govt should be here?

    1. Oh yes, very much Ramsey. It’s why we spun up the TOGI committee on fraud, for one – we need to get ahead of this or the government will have to do it for us. Thanks for stopping by!

  6. Have noticed that over the last few years Adobe has been acquiring a bunch of companies in the programmatic adtech space for their ‘marketing cloud’ service. It is interesting because it is almost orthogonal to their tools business and also could become a significant player.

  7. Interesting piece, John – makes sense to me. Although, people already don’t click on banner ads, just like they don’t order cans of Coke after seeing a TV ad. But many do respond so think viewthrough.

  8. My company found a novel way to repurpose our usage-based micropayment engine as an ad placement mechanism. It’s pure opt-in by the users. The system is connected at one end to customer loyalty/rewards programs and at the other end to ad servers. Not sure how to market it though. Would appreciate feedback and suggestions.

    Also curious if this is an instantiation of new type of ad banner John is referring to.

    A cloud-based “big-data” advertising platform which enables brands with loyalty programs to leverage members’ web/app usage to place ads strategically, and at the same time accelerate customer retention by offering points for active opt-in participation. The system is based on Kachingle’s existing production micropayment platform.

    More info here:.http://kachingle.com/site/crowdsourcedads

    Thanks for any comments.

    Cynthia Typaldos

  9. Programmatic display to date has been built on re-targeting
    data. By definition this is latent data modeled on previous actions or
    behaviors. Also, with the increased speed of the web 100-250ms is becoming too
    long to make the necessary computations prior to page load. Advanced systems require pre-calcs (similar to what AdWords does) to be done at the page or request
    level which means that these systems must be first-party (in addition to the
    privacy issues that are resolved being first party). It also requires more
    static data sets with higher signal to noise ration than what is used today in
    RTB and back-end data (outcomes) that need to be looped back in to aid in
    decision intelligence. This is not commonly part of RTB.

    I agree 100% with your vision of the future and have devoted a career to building it but what you describe is a targeting system that today more closely resembles site-side optimization technology doing things like landing page optimization (Adobe Test&Target), content recommendation (Outbrain) and cross-sell/up sell (Amazon) than what is known as “programmatic adtech” or “RTB” serving banners and represented in the LUMAscape.

    1. Interesting, thanks Jonathan. I find the first party issue to be the most intractable, as it favors locked in models and drives out innovation. I get your point about this not being common to RTB today, but argue with it only being driven by retargeting. I think that’s the main driver, to be sure, but it’s a model as well for any Request-Process- Response kind of system, no? Get the request, which includes a payload of contextual and other data, process against other data, return a request?

      1. First party data is important because the analytics and deep
        learning needed for high performance (requests) require that the source and timing of the data collection be dimensionalized. Without that it is impossible to expect that the application of that data can be used in the right context or timing. This doesn’t mean that there can’t be a service layer that provides the analytics (a new kind of Omniture for example) and one that provides the timing (a new kind of Ad Server). I see opportunity for vendor models and there is a great deal of innovation that can happen here. I have first hand knowledge it is.

        RPR is the future because this is exactly how the web functions. Every click is a request to a server that processes that request and responds with the matching resource. When we create real-time systems that can be embedded as applications within this layer they will function better than other systems that are really hacks (I would argue cookie based display is this type of system – but that’s for another time). You can see the advantage this provides Paid Search as the system operates on real-time user requests. It is pull vs. push – a very important distinction in understanding these systems. They are user controlled like the medium itself.

        Another great representative example of RPR is Waze where
        your timing and context is pulled in real-time. Waze processes that to respond with helpful and useful information depending on your location, your speed and other rules. Proof there can be innovation here — but also in the long run that Google will keep this market cornered as it’s core of web monetization.

Leave a Reply

Your email address will not be published. Required fields are marked *