Two candidates finish the same interview loop. One answers every question cleanly, has the polish of someone who's done this exact interview format before, and gives you nothing to worry about. The other stumbles on a system design question, recovers by asking a genuinely good clarifying question nobody expected, and leaves you a little unsure. Most hiring processes pick the first candidate. I think that's usually the wrong call, and it took me a few bad hires to notice why.
You're not hiring who's in the room
The candidate in front of you is a snapshot. What you're actually deciding is who they'll be in eighteen months, and a snapshot tells you almost nothing about the derivative — the rate at which someone is improving — unless you specifically go looking for it. Current level and slope are different measurements, and a hiring process optimised for the first will systematically miss the second.
This matters more than it sounds like it should because the two variables predict different things. Current level predicts how much ramp time you'll need. Slope predicts what the person is worth two years from now, and two years from now is where most of the value of a hire actually gets realised — nobody hires someone hoping they'll be exactly as useful in month twenty-four as they were in week one.
The candidate who's already polished for your interview format has usually optimised for exactly the wrong variable. Polish is a photograph. You're buying a trajectory.
I don't mean this as an argument for hiring anyone unpolished — polish correlates with plenty of things worth having. I mean that when polish and trajectory point in different directions, most processes default to rewarding the polish, because polish is legible in sixty minutes and trajectory isn't.
What actually signals slope
The clearest signal I've found isn't a question, it's what candidates do when they hit the edge of their knowledge. Someone hiring for current level wants to see the edge avoided — steer the conversation back to safe ground, keep the answer clean. Someone hiring for slope wants to see what happens at the edge itself: does the candidate get defensive, guess with false confidence, or do they say "I don't know, but here's how I'd find out" and then actually demonstrate that method in real time.
That third response is the one worth weighting heavily, and it's the one most loops are structured to punish, because it produces a messier-looking interview than the candidate who never left safe ground.
A second signal, less obvious: how someone talks about a past mistake. Not whether they can produce a mistake story — everyone's rehearsed one by now — but whether the story shows a specific, mechanical understanding of what they'd do differently, versus a generic lesson that could be bolted onto any failure ("I learned to communicate more"). The specific version is evidence of an actual feedback loop running in that person's head. The generic version is evidence they've learned to tell a story about having one.
A third: rate of change over their own history, if you can get at it. Someone who was mediocre eighteen months ago and is now genuinely good has just demonstrated they can do that again. Someone who was excellent eighteen months ago and is excellent now has demonstrated stability, which is worth something, but it's a different thing, and conflating the two is how a hiring process quietly starts optimising for pedigree over trajectory.
The tradeoff you're actually making
Hiring for slope is riskier in the way that matters to a hiring manager who has to justify the decision: it's harder to defend in a debrief, because "I have a strong instinct this person is on a steep curve" sounds weaker than "this person nailed every question." That asymmetry — slope is real but hard to defend, polish is often shallower but easy to defend — is exactly why most processes drift toward rewarding polish over time, even when everyone involved would say, if you asked them directly, that they care more about trajectory.
Correcting for that drift means being explicit about it: building the edge-of-knowledge and mistake-story signals into the loop on purpose, and being willing to make the harder-to-defend call when the trajectory evidence is strong. The candidates worth betting on are rarely the ones who gave you nothing to worry about. They're the ones who showed you, under pressure, exactly how they close the gap when they don't already know the answer — because that's the skill you're actually going to need eighteen months from now.