How do you word on-call and deploy questions to a peer engineer so the answer is a number, not an adjective?

## The problem with adjective answers Ask a peer engineer whether on-call is manageable and you will hear that it is manageable. Not because they are lying, but because the question offers a comfortable answer and social pressure supplies it. The same happens with 'are reviews fast?', 'is the codebase in good shape?' and 'is work-life balance respected?'. Every one of those has a default answer that costs the speaker nothing, which makes the reply worthless as evidence. The fix is not to be more aggressive. It is to ask for something a person can only answer by remembering. ## Two conversions that do the work **Opinion to count.** Replace the adjective with a quantity that has one true value. - 'Is on-call manageable?' becomes 'How many engineers are in the rotation?' - 'Do you deploy often?' becomes 'How many deploys went out last week?' - 'Are reviews fast?' becomes 'How long did your last change wait for its first comment?' - 'Do you get time for tech debt?' becomes 'What is the last piece of debt the team actually paid down?' **Policy to instance.** Replace the general period with a recent, specific one: last week, your last shift, your last change. A person answering about last week has to consult memory; a person answering about typical practice can answer from the recruiting narrative without noticing they are doing it. Combine both and the question becomes nearly unfakeable: 'How many people share the rotation, and what happened on your last shift?' ## Reading the numbers you get back For illustration, imagine two answers to the rotation question. One peer says six people share it and their last shift woke them once. Another says three people share it, one of whom joined recently, and their last shift woke them most nights. The second is not a scandal — a small team early in its life often looks like that — but rotation depth sets the floor on how often the pager returns to you, and depth of three with one person still ramping is a materially different job from depth of six. Deploy counts read the same way. Around 11 or 14 deploys in a week across a team says changes are small and the path is open; one release cut per week says changes batch and the feedback loop is long. Review turnaround is the third of the trio: a first comment usually within about 5 hours is a team that treats review as work; a peer who says their last change sat 31 hours before anyone looked has told you something no adjective would have. All of those figures are invented here to show the shape of a useful answer. The point is not the values; it is that a value exists at all, that you can compare it against what a second interviewer says, and that you can compare two companies on the same axis. ## The complement: the experience question Counts do not cover everything. Some of the most decision-relevant facts about a team have no metric: the tool everyone works around, the meeting that eats a morning, the part of the job that turned out to be better than the posting suggested. For those, the peer-round opener is 'What surprised you after you joined?' It works for three reasons. It is easy and pleasant to answer, so it rarely produces a defensive response. It cannot be answered from the recruiting narrative, because the narrative is exactly what the surprise contradicted. And it hands you a follow-up for free: whatever they name, ask whether it has changed since. ## Keeping the tone right A sequence of countable questions can start to feel like an audit, and a peer round is also scoring you. Three habits keep it a conversation. Ask for the number, then react to it like a colleague rather than a clipboard. Explain your interest in half a sentence when the question could sound loaded — 'I ask because I've been in a rotation of three and I want to know what I'd be walking into.' And stop at two or three; you are gathering enough evidence to decide, not producing a report. ## What good practice looks like Before the loop, write your questions as counts with a recency anchor already attached, so you are not converting them live. Afterwards, note the numbers immediately. The value compounds across a whole loop and across companies: three teams' rotation depths and review turnarounds beside each other is a real basis for a decision, and it costs one well-worded question per round.