We have hired senior engineers who aced the coding round and were a net drag within six months, and we have hired people who were merely solid on the keyboard and lifted the whole team. The difference was never typing speed. By the senior level, the code is the easy part to assess. The hard part is the judgment around it, and most interviews never look.
Can they decide without all the information?
Seniority is mostly the ability to make a reasonable call with incomplete data and own it. We give a deliberately underspecified problem, with constraints in tension, and watch what they do. Do they ask the question that matters most, name their assumptions, and commit to a direction? Or do they freeze waiting for someone to remove the ambiguity? The freeze is the tell. Real work never arrives fully specified.
- A messy, underspecified scenario, to see if they can move without certainty
- A past decision they got wrong, to hear whether they can own it without spin
- A disagreement we stage gently, to see if they can hold a position and also change their mind
- A walkthrough of how they would onboard, because seniors who cannot ramp themselves are juniors with a title
How they treat the people who are not in the room
We listen hard to how a candidate talks about past teammates, the junior who slowed them down, the manager who made a bad call. A senior who describes everyone else as the obstacle will be the obstacle here. The ones we want talk about what they could have done differently, and they give credit specifically. This is not a soft nicety. A senior sets the tone for everyone below them, and a corrosive one costs you people you cannot see leaving.
A brilliant engineer who makes the team smaller around them is a net loss. We measure seniors by the radius of the people they make better, not their personal output.
The coding round still matters, narrowly
We do still check they can write code, because we have been burned by the rare candidate who could talk for an hour and not produce a working function. But we keep it short and realistic, closer to debugging something half-broken than inventing an algorithm. We are confirming competence, not ranking it. Above the bar is above the bar; we do not pay extra for someone who can invert a tree faster.
What we stopped doing
We dropped the trivia and the puzzle that has one clever trick. They selected for people who had recently practiced interviews, not people who had recently shipped things that mattered. We replaced them with conversations about real tradeoffs the candidate has actually faced, and the signal got dramatically better. The best predictor of doing the job well turned out to be talking honestly about having done it before, including the parts that went sideways.