I am currently watching the progress bar on a SQL query that has been running for 35 minutes, and while the blue line crawls toward the right, I am practicing my signature on the margin of a printed report. I’ve done it 25 times now. Each loop of the ‘W’ feels a bit more certain, a bit more like a person who actually stands behind their words, which is a stark contrast to the document I’m waiting for. This query is supposed to tell us if we should launch the new interface. We already know the answer. The beta testers loved it, the developers are proud of it, and the old version looks like a fossil from 2005. Yet, here we are, burning 555 dollars of server time to find a statistical reason to do what we already decided to do three weeks ago.

We call this being ‘data-driven,’ but let’s be honest: it’s actually a sophisticated form of hiding. We are using numbers as a human shield. If the launch fails and we can point to a p-value of 0.05, we aren’t incompetent; we were just following the data. It’s the ultimate absolution of the modern professional. We have traded the terrifying, exhilarating weight of leadership for the cold, unblinking safety of the dashboard. I watched a room of 15 stakeholders spend 105 minutes debating the shade of a notification dot yesterday. We had 45 slides of eye-tracking heatmaps. Not one person said, ‘I think this looks better.’ Instead, they said, ‘The data suggests a 0.5 percent increase in peripheral gaze retention.’

It is a lie. We aren’t optimizing; we’re stalling because we are afraid to be wrong. I know this because I’ve been the one building the shield. I’ve spent 255 hours of my life aggregating metrics that served no purpose other than to make a Vice President feel like they weren’t making a guess. But every great move is a guess. Every single one. Data can tell you where you’ve been, and it can tell you if the bridge you just built is currently upright, but it cannot tell you if the destination on the other side is worth the trip. That requires a soul, or at least a gut, and both of those things are currently being regulated out of the corporate ecosystem.

The Crossword Analogy

My friend Alex W. knows this better than anyone. Alex constructs crossword puzzles-not the easy ones you find on the back of a flight magazine, but the 15×15 behemoths that make you question your own literacy. If you look at the data of crossword construction, there are certain words that appear with grueling frequency. ‘Eerie,’ ‘Area,’ ‘Opal.’ These are the ‘safe’ words. They have a high vowel-to-consonant ratio and fit into any grid. A data-driven crossword would be a repetitive nightmare of these 5 words over and over again. It would be statistically perfect and utterly soul-crushing.

Alex doesn’t work that way. He’ll spend 75 minutes trying to fit a single colloquialism or a weirdly specific cultural reference into a corner because he knows it will give the solver a moment of genuine delight. There is no metric for ‘delight’ that a spreadsheet can capture without stripping it of its context. He takes the risk. He puts his name on the grid. He signs it. When was the last time any of us truly signed our work without a footnote citing a data source? We’ve become a culture of ghost-writers for our own algorithms.

I remember a specific project 15 months ago. We were redesigning a checkout flow. The data was screaming at us to add more ‘nudges’-those little countdown timers and ‘5 people are looking at this’ pop-ups that make you feel like you’re in a burning building. The metrics said they worked. Conversion would go up by 5 percent. But sitting in that room, looking at the wireframes, I felt a physical sense of grime. It was ugly. It was manipulative. It treated our users like lab rats rather than humans. I pointed this out, and the response was a blank stare from 5 different people. ‘But the data says it’s better,’ they repeated. They weren’t even looking at the screen; they were looking at the bar chart.

We did it anyway. Conversion went up, just like the numbers predicted. And then, 5 months later, our brand sentiment scores plummeted. People didn’t trust us anymore. They felt bullied. You can’t A/B test the slow erosion of trust until it’s already gone. By then, the people who made the ‘data-driven’ decision have already moved on to their next promotion, leaving the wreckage for someone else to clean up. This is the danger of the localized maximum. We optimize for the next 15 minutes and ignore the next 15 years.

Reclaiming the Signal

I think about the tools we use to navigate this mess. We are constantly seeking clarity, but we look for it in the noise of the crowd rather than the stillness of our own minds. People often turn to brain honey when they realize that the external noise is just a distraction from the internal signal. It is about reclaiming that mental space where actual decisions happen-the kind of decisions that require a person to stand up and say, ‘I believe this is the right path,’ regardless of what the 225 rows of Excel say. It’s about getting back to a place where we trust our own expertise as much as we trust a tracking pixel.

There is a specific kind of exhaustion that comes from justifying the obvious. It’s a cognitive tax we pay for the privilege of working in a hierarchy. I’ve seen 45-year-old men with three degrees stutter in front of a dashboard because the ‘trend line’ went down by 5 percent in a single afternoon. They forget that the trend line is just a shadow of human behavior, and shadows are notoriously fickle. They forget that they were hired for their judgment, not for their ability to read a graph that a moderately clever golden retriever could interpret.

I’m not saying we should burn the servers and go back to divining the future through goat entrails. Data is a tool. It’s a hammer. But if you spend all day measuring the hammer instead of hitting the nail, you aren’t a carpenter; you’re a bureaucrat. And we have far too many bureaucrats in creative clothing. We have replaced the ‘creative director’ with the ‘optimization lead,’ and the result is a world that feels increasingly like a polished, frictionless airport lounge. It’s efficient, yes. It’s safe. It’s also entirely forgettable.

The ‘Aha!’ Moment

Alex W. once told me that the best part of a crossword is the ‘aha’ moment-the split second where the brain leaps across a gap that logic couldn’t bridge. You can’t calculate that gap. You have to feel it. You have to know the audience. You have to be willing to fail if the clue is too obscure. That willingness to fail is the only thing that creates something worth succeeding. If you are ‘data-driven’ to the point of risk-elimination, you aren’t actually succeeding; you’re just surviving at a very high resolution.

I look down at my signature on the legal pad. It’s getting better. The 35th version is the best one yet. It’s bold, it’s messy, and it’s mine. The SQL query finally finishes. The results are exactly what I expected. The new design is better by a margin of 15 percent. I could take this to the meeting and spend 55 minutes explaining the methodology, the cohort selection, and the standard deviation. Or, I could just show them the design and say, ‘This is what we’re doing because it’s better.’

I’ll probably do both. I’m still a coward in a cubicle, after all. But maybe I’ll leave the data slides for the very end, as an appendix that no one reads. Maybe I’ll spend the first 25 minutes talking about the feeling of the interface, the rhythm of the user journey, and why it matters that we make something that doesn’t feel like a spreadsheet. We have to start reclaiming the right to be human in the face of the algorithm. We have to start signing our work again.

It’s a strange thing, this addiction to certainty. We crave it like a drug, even though every major breakthrough in human history came from a state of profound uncertainty. You don’t get to the moon with a p-value. You get there with a dream and 55,000 pounds of rocket fuel and a guy named Neil who had to make a split-second decision to land the thing manually because the computer was failing. If they had been ‘data-driven’ in the modern sense, the Eagle would have stayed in orbit forever, waiting for more telemetry.

So here is my challenge, to myself as much as to you: For the next 15 days, try to make one decision based on nothing but your own expertise. Don’t look at the analytics. Don’t run a survey. Don’t ask for a second opinion from a machine. Just look at the problem, look at the solution, and if your heart beats 5 beats faster when you see the answer, do it. The world won’t end. The numbers might even go down for a minute. But at least you’ll be able to look at the result and see your own signature staring back at you, instead of a ghost.