Why marketers have to stop using yesterday’s rules to manage today’s advertising technology
There is a question we get from clients all the time: “Can you give us a list of the websites where our ads appeared?” Absolutely. We are happy to provide it. Transparency matters, and advertisers should know where their advertising is appearing.
But what often happens next is much more interesting. Someone starts going through the list. What kind of website is this? What industry does it represent? Why would our customer be there? Does this site have the right affinity? Should we exclude this one? Should we serve more impressions on that one?
I understand the instinct because marketers have been trained to think this way for decades. We selected publications. We selected programming. We selected websites and categories. We built demographic profiles and affinity groups. We decided where we believed our customers would be, and then we bought media in those places. In that world, where the ad appeared was an important part of why the ad appeared.
But that isn’t necessarily the world we are operating in anymore. When advertising is driven by real-time behavioral data, AI and purpose-built algorithms, the website can be the place where the right person happened to be when the opportunity to reach them occurred. That is a very different idea, and I have a story that explains it better than almost anything else I could say.
The Multi-Thousand-Dollar Bottle of Wine
Several years ago, we were working with a luxury wine importer. These weren’t inexpensive bottles of wine. Some of the inventory sold for thousands of dollars per bottle. One of those very expensive bottles sold at approximately 4:00 in the morning to someone in Oklahoma who was using a solitaire app.
Think about that for a moment from the perspective of a traditional media planner. Imagine walking into the client’s office before the campaign and saying, “We have a great idea. We’re going to advertise one of your most expensive bottles of wine on a solitaire app at 4:00 a.m. in Oklahoma.” We probably would have been shown the door, and understandably so.
Almost nothing about that placement would have passed the traditional marketing test.
The content wasn’t about wine.
The environment didn’t communicate luxury.
The time of day looked wrong.
The geography probably wouldn’t have been the first place someone planning a luxury wine campaign would have chosen. If we had been evaluating the opportunity based entirely on the placement, we could have found a dozen reasons not to serve that ad.
But that is where the bottle sold. That story has stayed with me because it illustrates something fundamental about how marketing is changing: the placement didn’t make sense, but the person did.
The technology wasn’t trying to find a website that looked like the perfect environment for an expensive bottle of wine. It was trying to identify the right consumer based on the signals available to it and reach that consumer when an opportunity existed. The solitaire app was where that person happened to be.
People Don’t Live Inside Marketing Categories
This is one of the problems with the way marketers have traditionally thought about audiences. We create personas. We create demographic groups. We create affinity categories. We decide what our ideal customer probably reads, watches, earns, drives, buys and does in their spare time. Those tools were useful because, for a long time, they were some of the best information available to us.
But real people don’t live inside our marketing categories.
People who buy expensive wine also play solitaire.
Someone considering a new HVAC system still checks sports scores.
Someone looking for a tax professional might be reading a recipe.
Someone considering cosmetic surgery might be checking the weather.
Someone thinking about joining a gym may be reading financial news.
A person’s entire digital life does not suddenly become about the product they are considering.
That seems obvious when we say it out loud. Yet marketers routinely look at placement reports and try to infer the value of the consumer from the subject matter of the page the consumer happened to be viewing. Those are two very different things.
Where We Found Someone Is Not Why We Selected Them
That distinction becomes increasingly important as advertising becomes more intelligent. A site list tells us where an impression was served. It does not necessarily tell us why that person was selected. Those used to be much closer to the same question. Today they can be almost completely separate.
If our technology is identifying behavioral patterns and signals that suggest someone is becoming relevant to a business objective, we don’t necessarily want to wait until that person arrives at a website conveniently labeled with the advertiser’s industry. By then, the intent may already be obvious to everyone. The opportunity is to recognize people earlier.
That means the intelligence increasingly sits in the behavior of the consumer rather than in the category of the webpage. The website may be where we found the person. It isn’t necessarily why we selected the person.
This Doesn’t Mean Placement Doesn’t Matter
There is an important distinction here.
Brand safety matters.
Fraud matters.
Inventory quality matters.
Suitability matters.
There are environments where a brand should not appear regardless of how attractive an individual impression might otherwise be. Those are legitimate controls, and marketers should have them.
But brand safety is very different from looking at a perfectly acceptable website or app and deciding, “I don’t think our customer would be here.” That is where human intuition can unintentionally begin working against the intelligence we asked the technology to provide.
If we had decided that people capable of buying multi-thousand-dollar bottles of wine shouldn’t be playing solitaire at 4:00 in the morning, our assumption would have been perfectly reasonable. It also would have been wrong.
We Keep Trying to Turn AI Back Into a Media Plan
This is where I think marketing is struggling with AI in a much larger way. We say we want AI. We want machine learning. We want predictive intelligence. We want real-time optimization. We want systems capable of analyzing more signals and making decisions faster than humans possibly can.
Then the system makes a decision we wouldn’t have made, and our first instinct is often to make it behave more like us.
We want AI, but we want it to choose the websites we would have chosen.
We want algorithms, but we want the audiences to look like the affinity groups we already understand.
We want machine learning, but we want to manually approve the conclusions.
In other words, we want new technology operating inside old assumptions. At some point, those two ideas collide. If the only decisions we allow an intelligent system to make are the decisions we would have made ourselves, we have eliminated much of the reason for using the intelligence in the first place.
Marketing Has Always Had Trouble Letting Go
This isn’t the first time we have done this. Last-click attribution may be the clearest example. More than a decade ago, Google was already publicly arguing that marketers needed to move beyond last-click attribution. The problem was straightforward: giving all the credit to the final interaction ignored everything that happened before it.
And yet marketers still struggle with this today. A customer can see advertising, encounter a brand repeatedly, visit a website, leave, return later, conduct a branded search and finally convert through Google. The CRM records Google, and suddenly Google “created” the customer. It didn’t necessarily create the customer. It recorded the last measurable step.
We know customer journeys are more complicated than that. We have known it for years.
But last click gives us something human beings love: a clean answer.
Site lists can do the same thing. This website looks right. That website looks wrong. This affinity makes sense. That one doesn’t. It feels like control, and sometimes what feels like control is really just familiarity.
The Marketplace Doesn’t Care What Makes Us Comfortable
This may be the bigger lesson. Marketing organizations can take years to change their thinking. The marketplace doesn’t. Consumers changed whether we were ready or not. Privacy changed. Search changed. Media consumption changed. The customer journey changed. The amount of behavioral information available to sophisticated systems changed. And our ability to identify patterns inside those signals changed dramatically.
We can continue evaluating this new environment using frameworks created for an earlier era. But doing that doesn’t force consumers to behave according to those frameworks. It just limits what we are willing to see.
That is why I think one of the hardest parts of AI isn’t actually adopting the technology. It is letting go of the assumptions the technology has made obsolete.
Control Hasn’t Disappeared. It Has Moved.
None of this means marketers should turn everything over to a black box and stop asking questions. Quite the opposite.
Humans should define the business objective.
Humans should determine what success looks like.
Humans should establish brand standards, geographic boundaries, budget parameters and unacceptable environments.
Humans should demand transparency.
Humans should challenge performance. Humans should make sure the technology is solving the right business problem.
But there is an important difference between establishing guardrails and micromanaging individual decisions. The role of the marketer is increasingly to tell the system what we are trying to accomplish and what boundaries it must respect, while allowing the technology to discover patterns within those boundaries that we may never have found ourselves.
That is still control. It is just control at a different level, and I think that distinction is going to become increasingly important as AI becomes more deeply embedded in marketing.
Sometimes the Surprise Is the Point
I sometimes wonder what would have happened with that bottle of wine if we had been given a traditional placement plan to approve before the campaign. Luxury publications would have made sense. Wine publications, of course. Food and travel, probably. Affluent lifestyle content would have seemed like an obvious choice. But a solitaire app at 4:00 a.m. in Oklahoma? We probably would have deleted it, and we might have deleted the customer along with it.
That is the part marketers need to think about. When an intelligent system discovers something that doesn’t fit our assumptions, our first reaction shouldn’t always be to correct the system. Sometimes we should question the assumption.
Because the real promise of AI isn’t that it can execute our existing marketing ideas faster. It is that it can sometimes see patterns we wouldn’t have thought to look for at all.
That requires a different relationship between marketers and technology. Not blind trust. Not unlimited freedom. But enough trust to allow the system to do what we built it to do.
We have spent decades deciding where customers should be. Today, increasingly, we can identify where the right customer actually is. Sometimes that’s exactly where we expected them. And sometimes it’s a solitaire app in Oklahoma at 4:00 in the morning.
The placement didn’t make sense. The person did.
By Carole Lawson
Co-Founder & Chief Data Officer, MarketStorm