Data Beats Gut, but Only When It Supports the Call You Own

You've been in the review where someone waves a dashboard like a court order. The number said go, so we went. That's not what strong teams do with data, and this week three very different groups made the same point clear. A Super Bowl coach, a billion-dollar telescope, and a Google traffic study all landed on one rule. Let me catch you up.

The number is a witness, not the judge

Seahawks head coach Mike Macdonald calls himself a "psycho data guy" who needs numbers and tendencies. He also says you don't have to do what the numbers say. Both things are true at once. He reads the model, then weighs how the game feels and what he knows about his own team.

Last season, facing 4th-and-1, his model said go for it. He kicked the field goal instead and lost that game. Six months later he had a Super Bowl ring. The lesson is not that data is wrong. It's that the person who owns the call still owns the call.

For your team, that means the metric is evidence you bring to the decision, not the decision itself. When a designer defends a choice, "the data supports this" is a stronger position than "the data made me."

The data still needs a human to say "we're ready"

The Vera C. Rubin Observatory just started a ten-year survey of the southern sky. Its camera shoots 3,200 megapixels every 40 seconds and will pour out terabytes a night. That's about as data-rich as a project gets.

And yet a person decided when to start. UW astronomer Zeljko Ivezic said the call to begin the survey came after a careful review of technical readiness, data performance, and scientific validation. The instrument doesn't get to declare itself ready. A team looks at the readouts and makes the judgment.

Same move for you. All the telemetry in the world still needs someone to read it, weigh it against what they know, and put their name on "ship it."

Small, aimed changes beat sweeping ones

Google Research ran a six-month experiment in 10 US cities, nudging drivers around known bottlenecks. They changed routing for under 2% of trips. The payoff was a median 2% bump in speed on targeted roads and thousands of tons of CO2 saved per city per year.

They didn't reroute the whole city. They picked roughly 100 congested segments per city based on real history, then moved a sliver of traffic. Small, aimed, measured.

That's a template for a roadmap review. You don't need a redesign to prove impact. Find the few spots where friction stacks up, change those, and measure the lift. A 2% gain on the right screen beats a splashy overhaul you can't defend.

Data earns its keep by getting specific

Google's heat team went from neighborhood averages to building-level rooftop reflectivity across 50-plus cities. That jump in resolution is the whole point. General climate data told planners a city was hot. The 30-cm maps told them which roofs to fix first.

That granularity guided real calls, like cool-roof ordinances and adaptation plans. Their modeling suggests targeted planning could cut extreme urban heat by up to 0.5°C. Vague data gets nodded at. Specific data gets acted on.

Bring that to your next review. "Engagement is down" is a shrug. "This step loses 30% of new users on mobile" is a decision waiting to happen.

The deep cut

All three of these winners kept a human on the hook. The coach owned the field goal. The astronomer owned the start date. Google framed its work as tools for planners, not a machine that decides for them. So when a designer walks into your review, don't let them hide behind a chart. Ask them to state the call, then show the data that backs it. The number is their witness. They're still the one testifying. A team that can say "here's my decision and here's the evidence" will beat a team that says "the dashboard told us to" every time.

Three questions for your team

1. On our last big call, did a person own it and use the data to support them, or did we let the number decide and skip the judgment?
2. Where are our top three bottlenecks, the equivalent of Google's 100 segments, and what would a small aimed fix there actually move?
3. Is our data specific enough to act on, like a roof-by-roof map, or are we still staring at neighborhood averages and calling it insight?