Back in March, I listened to a seven-hour interview with Saining Xie, Co-founder and Chief Scientist Officer of AMI Labs. One phrase stuck with me long after the interview ended: Research Taste.
He described research taste as the ability to identify important problems, distinguish signal from noise, and develop intuition for which directions are worth pursuing. What struck me wasn't the definition itself, but the idea that beyond technical skill, there is another layer of judgment that determines whether someone consistently works on the right things.
I kept thinking about that phrase because it reminded me of something I'd been observing throughout my own career, although I had never given it a name.
Over the past few years, I've worked closely with economists, data scientists, machine learning engineers, product managers, and marketers. Many of them were far stronger than I was in their respective domains. Yet one pattern kept showing up: the most sophisticated solution was rarely the most successful one.
Some of the highest-impact systems I've seen weren't the most advanced models, the most elegant experiments, or the most complex optimization frameworks. They were the systems that people actually trusted, adopted, and used to make decisions.
That observation became even more obvious as AI capabilities accelerated. We gained access to better models, more data, more automation, and increasingly sophisticated ways to optimize nearly everything. Yet the same question kept appearing underneath all the technical progress:
What actually creates value?
Eventually, I realized I was describing a form of judgment. Not research taste, but something adjacent to it.
Maybe marketing has its own version.
Maybe it's what I would call marketing taste.
To me, marketing taste is the ability to distinguish between what is technically impressive and what actually drives business outcomes. It's knowing when a sophisticated model is worth the added complexity and when a simple heuristic is sufficient. It's recognizing the difference between improving a metric and improving a decision.
At its core, marketing taste comes down to a deceptively simple question:
Does this actually help the business grow?
Research taste asks whether something is true. Engineering taste asks whether something can be built. Product taste asks whether people will use it. Marketing taste asks whether it creates value.
The distinction sounds subtle, but I've found it leads to very different decisions.
One of the easiest traps in growth is optimizing what is measurable instead of what is meaningful. Over time, attribution models become more sophisticated, prediction models become more accurate, dashboards become more detailed, and metrics become increasingly granular. Yet business outcomes don't always improve alongside them.
I've seen teams spend months refining a model while creating almost no meaningful change in how decisions were made. The issue usually wasn't the quality of the model itself. The issue was that nobody stopped to ask whether the improvement would actually influence behavior.
Because growth rarely comes from better reporting. Growth comes from better decisions.
This is where I've come to see the difference between research taste and marketing taste.
Research taste is concerned with questions like: Is the methodology correct? Is the model statistically sound? Does the experiment meet scientific standards? These are important questions, and many of the systems I've worked with would not exist without people who care deeply about them.
But marketing taste introduces a different set of concerns. Will anyone trust this output? Will anyone use it? Will it influence a decision? Will it ultimately improve outcomes?
I've seen technically brilliant solutions fail in practice-not because the math was wrong, but because the solution didn't fit how humans actually make decisions. A model can be right and still be ineffective. In many cases, adoption matters just as much as accuracy.
This is also why I think AI is making marketing taste more important, not less.
A common assumption is that as AI systems become more capable, human judgment becomes less necessary. My experience has led me to the opposite conclusion. AI can generate more recommendations, predictions, optimizations, content, and decisions than ever before. Information is no longer the bottleneck.
Judgment is.
AI can optimize whatever objective we give it. But someone still needs to decide whether the objective itself is the right one. Someone still needs to determine which signals deserve attention, which trade-offs matter, and whether the outcome is creating real value.
As systems become more capable, the quality of our judgment matters even more.
This is also why I've never fully believed that human-in-the-loop is merely a temporary phase before full automation. In many real-world marketing environments, the goal isn't to eliminate human judgment. It's to augment it.
Marketing operates in messy environments full of uncertainty, incomplete information, competing incentives, and constantly shifting business priorities. There is rarely a single objectively correct answer. The best systems I've seen combine machine intelligence, human judgment, and business context. Not because humans are better calculators than machines, but because value creation often requires navigating ambiguity rather than simply optimizing an objective function.
Looking back, I realize my own understanding of growth has evolved several times. Early in my career, I thought growth was mostly about optimization. Later, I thought it was about measurement. Then experimentation. Then automation.
Today, I still believe all of those things matter. But I no longer think they're the highest level of the game.
The highest level is judgment.
It's knowing what matters and what doesn't. It's knowing which signals deserve attention and which can be ignored. It's knowing when to trust a model and when to challenge it.
Marketing taste is not about knowing how to optimize a system.
It's about knowing what is worth optimizing in the first place.
And in a world increasingly powered by AI, that may become one of the most valuable skills we have.