The Dashboard Is Not the Customer

By Chris Marine, Founder/CEO of Campfire Consulting

Marketing has spent the better part of two decades getting very good at measuring what happened. We know what people clicked, what they watched, what they bought, what it cost to reach them and, depending on the model, which channel gets credit for the outcome. None of that is bad. In fact, more accountability has made media better. But somewhere along the way, we started treating what is easiest to measure as if it were the same thing as what matters most.

That tension came up recently in a conversation I had with Dr. Ram Singh, CEO and Founder of EtaKor. Ram has spent much of his career in advanced analytics and measurement, and he made a distinction that has stuck with me. He described many of the metrics marketers rely on as “convenience metrics.” They are useful, sometimes very useful, but they exist in part because they are available to us. The fact that something can be captured cleanly in an ad log does not automatically make it the best way to understand whether marketing is actually working.

ROAS is probably the most obvious example. It’s a helpful metric. It’s also one we ask to do far too much. We know the spend side of the equation with certainty. The return is more complicated. Did the media create the sale, accelerate a decision that was already underway or simply get credit for demand that already existed? What about the person who saw a campaign several times, did nothing, and remembered the brand three months later when a need or want finally emerged? That value is real, but it rarely fits neatly into the reporting window.

This matters because the industry is rapidly automating more of the decision-making around media. AI can find patterns across enormous volumes of data, adjust bids faster than any person and surface relationships a human analyst might never see. That’s progress for sure. What worries me is the temptation to assume that better pattern recognition means human behavior itself has become predictable.

It hasn’t.

People are inconsistent by nature. Someone can fit every characteristic of a high-intent customer and still not buy. A message can reach the right person at the “right” time and still land flat because they are distracted, tired or simply not in the mood to hear it. Another message can appear to do very little in the short term while quietly making a brand more familiar over time.

Ram talked about the signals that often get discarded as noise because they are harder to explain. His argument was that some of those weak signals may actually hold the most interesting information. I think that’s where marketers and advertising practitioners should be spending more energy.

We should be asking what our dashboards cannot answer cleanly. How long does it take for a brand to become familiar enough to be considered? What role does the sequence of creative play? Is a campaign creating new demand or simply capturing people already close to buying? When does a message need more time versus more money? Those are harder questions than whether a click-through rate went up 12 percent, but they are closer to how people actually experience brands.

It also changes how we think about measurement itself. Too often measurement is brought in after the strategy, the creative and the media plan are already finished, almost as an auditing function designed to prove whether the plan worked. We would be better served by bringing it into the room earlier, using it to help shape the questions we are trying to answer before a dollar is spent.

The most interesting part of my conversation with Ram was not really about AI or analytics at all. It was about judgment. When I asked what part of human behavior he doesn’t believe an algorithm can replicate, his answer was critique. The ability to look at an answer, poke holes in it and introduce a possibility the model didn’t consider.

That’s where I think the role of the strategist becomes more important, not less.

Machines will continue getting better at finding patterns. Our job shouldn’t be to compete with them at that. Our job is to understand what those patterns mean, where they may be misleading us and what context is missing from the answer.

As Ram put it, “measurement is a vehicle, not an end.”

The dashboard should help us understand people better. We should be careful not to turn people into something that only makes sense inside the dashboard.

This article was inspired by my conversation with Dr. Ram Singh, CEO and Founder of EtaKor, on the Responsibly Human podcast. Listen to What the Data Misses About Being Human on Apple Podcasts.