SQL, R, Python — none of it is what makes a good analyst.

I’ve hired and managed many data scientists and analysts, and the strongest predictor of who’d actually be great at the job was never their skill set. It was their talent: the stuff that comes naturally to a person and doesn’t for most people. Skills can be taught in a weekend course. Talent can’t.

I learned that distinction from First, Break All the Rules (Buckingham & Coffman), which is still the best management book I’ve ever read. The book was built on Gallup’s research into what separates great managers from average ones, supported by data from thousands of companies. Here’s how I’ve applied it to analytics teams specifically.

Hire for talent, not credentials. A resume full of the right tools tells you someone can learn. It doesn’t tell you whether they have the curiosity to chase down why a number looks wrong, or the stubbornness to keep testing a hypothesis after the first two answers were dead ends, or the attention-to-detail to notice that numbers on different outputs that should be similar are very different. Those aren’t skills you train into someone.

Give them real autonomy. The best analysts I’ve managed did their best work when I told them the destination and then got out of the way and let them choose the route. Micromanaging is frustrating for all concerned, and robs you of the unique approach your direct report might have taken.

Set the target, then measure the target. Vague goals produce vague work. The fastest way to derail a good analyst is to leave them guessing what “good” looks like for this project.

Protect the learning curve. This field moves fast enough that an analyst who stops learning starts falling behind within a year or two. The best teams I’ve built treated ongoing learning as part of the job, not an extracurricular.

The analysts who leave first are generally not the weakest ones. They’re the ones who feel like their talent isn’t being used, and that’s a “you” problem, not a “them” problem.