Learn

In a behavioral round, say the decision that was yours

16 October 2026 · Behavioral: your part of the story

The interviewer asks about a time you improved something that kept breaking. You tell a good story, it is true, and every sentence starts with "we". Then comes the follow-up: "What was your part in that?" You realise you have spent three minutes describing the team, and you start the story again from the middle.

The situation

A behavioral round is scored on what one person did, decided and learnt. The team matters, but the interviewer cannot hire the team. When the story is told in the plural, they have nothing to write down except that a project happened near you. Prep guides repeat the rule "use I, not we", and it is right as far as it goes. The harder part is that swapping pronouns does not help if you have not found the decision you actually made.

A recent post by an engineer describes this moment well: an answer that was a true number and nothing else, followed by a question about which part of the system they were most afraid of. That question is the one to prepare for.

What goes wrong

Here is an invented story, the kind many engineers could tell:

We had a nightly import that failed about once a week. We looked into it and found it was timing out on a slow partner endpoint. We added retries and better monitoring, and after that it was much more stable. The team was really happy with it.

Four sentences, five uses of "we", and no decision. Retries and monitoring are what anyone would add. The listener cannot tell whether you diagnosed the timeout, wrote the retry, argued against it, or reviewed someone else's change.

Now put one probing question to it: "Why retries, and not something else?" If the honest answer is "that was the obvious fix", the story has no centre. If the honest answer is "I pushed back on plain retries because the import was not idempotent, so a retry after a partial write would double some rows", you have found the story. It was always there; the plural hid it.

The test is simple. A decision is something where a reasonable colleague could have chosen differently. "We added monitoring" fails the test. "I chose to make each batch idempotent before adding any retry" passes it, because someone could have argued for retries first.

What to say out loud

Keep the team in the story, and make your own move visible in one sentence:

The team agreed the import needed retries. My part was noticing that a retry after a partial write would duplicate rows, so I proposed making each batch idempotent first, keyed on the partner's record id, and I wrote that change. After it shipped, a failed night could simply be rerun, and we stopped doing manual clean-ups the next morning.

Three things happen there. The team gets credit for what it decided together. Your decision is named, with the alternative you argued against. The effect is something an observer could check: a rerun replaced a manual clean-up. Avoid inventing a percentage you never measured. A plain, observable change is stronger than a number you cannot defend in the follow-up.

If the interviewer pushes ("what would you do differently?"), have one honest answer ready about the cost of your decision, for example that the idempotency key added a lookup per row and you had not checked that it scaled.

The exercise

1. Write down one story you plan to tell, as you would say it today, without editing. Underline every "we".
2. For each underlined sentence, ask: could a reasonable colleague have chosen differently here, and was it my call? Keep only the sentences where both answers are yes. If none survive, pick a different story.
3. For the decision that survives, write the alternative you rejected and why, in one sentence.
4. Write the effect as something an observer could see, not as a feeling. "The team was happy" does not count.
5. Tomorrow, retell the story aloud from memory without your notes, in under two minutes. If the decision sentence has drifted back into "we", the story is not yet yours. Do it again in three days.

Where this fits in Gadfly

Gadfly has four rooms: algorithms, system design, behavioral and code review. The behavioral room keeps a bank of your stories, separates the competencies you claimed from the ones the story actually earned, and asks you to retell a story later, which is the last step of the exercise above. Its report is a set of observations made by rules over what you said, not a hiring prediction.