0%
Behavioral Interviews for Software Engineers
The Common Question Families
Senior and Staff Signals
Showing Impact with Numbers
Impact is one of the signals that appears on almost every behavioral rubric, and it is the one where candidates lose the most points for the least reason. Many engineers did work with real, measurable effects and then describe it as it went well or the team was happy. This lesson explains why numbers matter so much to scorers, how to find them, how to choose the right ones, how to present them credibly and what to do when the numbers are missing.
What a scorer can write down
An interviewer taking notes needs evidence they can record and defend in the debrief. It went well cannot be defended. p99 latency fell from 900 to 300 milliseconds, and checkout conversion rose by about two percent the following month, can be. Numbers turn a claim into evidence. They let the interviewer write a specific note, and they let the committee compare candidates on something more solid than impressions.
Numbers signal that you measured
A number in a story says more than the size of the effect. It shows that you thought to measure before and after, which is itself a behavior the rubric values. Engineers who measure tend to make better decisions, catch regressions and know when to stop. When a candidate gives no numbers, the interviewer is left to wonder whether anyone measured anything, which weakens even a genuinely strong story.
Numbers set the scope
Numbers also tell the interviewer how big the work was. A latency fix that saved two milliseconds on an internal tool and one that cut checkout time for millions of users sound identical without numbers. Scope is central to leveling, so a story without numbers can be read at a lower level than it deserves simply because the interviewer cannot see its size.
The cost of vague results
Vague results are one of the most common reasons strong stories score a two instead of a three. The actions may be excellent, but a result described as it improved a lot leaves the scorer unable to rate impact above the bar. Candidates often do not notice this in practice, because they know the result themselves and do not realise the interviewer cannot see it.
Numbers make follow ups easier
A result with a number invites a useful follow up: how did you measure that, what was the baseline, what else changed at the same time. These questions let you show rigor. A vague result invites a less useful follow up: can you quantify that, which puts you on the back foot and often leads to an uncertain estimate. Leading with the number puts you in control of the conversation.
Numbers are not only for results
Numbers belong throughout a story, not only at the end. In the situation, they set the stakes: three billing incidents a month, forty percent of pages were noise. In the actions, they show reasoning: profiling showed eighty percent of time in one query. In the result, they show the effect. A story with numbers in all three places reads as the work of someone who works from evidence.
Numbers without arrogance
Some candidates avoid numbers because they feel like bragging. Numbers are the opposite of bragging. A boast is a claim about yourself, such as I am very good at performance work. A number is a fact about the work that the interviewer can judge for themselves. Stating facts plainly, and letting the interviewer draw the conclusion, comes across as confident and modest at the same time.
How many numbers a story needs
A good story usually has two to four numbers: one or two in the setup to set the stakes, perhaps one in the actions to show the reasoning, and one or two in the result. More than that tends to overwhelm the listener, who cannot hold many figures in mind while following a story. If you have more, keep the extras in your follow up notes for when the interviewer asks.
Numbers and seniority
The kind of number expected changes with level. Junior engineers are credited for task level numbers such as test coverage or build time on their own component. Senior engineers are expected to connect technical numbers to user or team outcomes, such as conversion or on call load. Staff engineers are expected to connect work to business or organisational outcomes, such as cost across many teams or a strategic launch date. The level lesson covers this in depth.
Where this lesson fits
This lesson builds on the story structure and story bank lessons. The result step of every story should already contain a number, and every bank entry should have a numbers section. Here we look more closely at how to find, choose and present those numbers well, and what to do when they are hard to find.
Numbers travel through the debrief
An interviewer does not only score you. They summarise you to a committee, often days later, alongside several other candidates. Numbers survive that journey far better than adjectives. A committee member who reads cut checkout p99 from 900 to 300 milliseconds understands the claim immediately, while one who reads strong performance work has to trust the interviewer judgment. The more of your evidence that survives the debrief intact, the more consistently your stories are credited.
Numbers protect against a harsh interviewer
Some interviewers score strictly, and some are sceptical of candidates who seem polished. A measured result is the best defence against both. A strict interviewer can still dock you for weak actions, but cannot dismiss a clear before and after. A sceptical one can probe the number, but if it holds up under probing, their scepticism turns into confidence. Vague results leave you entirely at the mercy of how generous the interviewer feels that day.
The difference is often one sentence
The gap between a vague result and a measured one is usually a single sentence. Candidates who score poorly on impact often did impactful work and simply ended the story with it worked out well. Adding one sentence with a baseline, an outcome and a unit is the cheapest improvement available in behavioral preparation. Before any loop, check that every story in your bank ends with that sentence, and practise saying it.
Numbers help you as much as the interviewer
Finding the numbers also improves your own understanding of the story. Candidates often discover, when they go looking, that the result was bigger than they remembered, or that one action mattered more than another. Sometimes they discover that a story they thought was strong had little measurable effect, which is valuable to know before telling it to an interviewer. The search for numbers is a test of the story itself.
Practise saying numbers aloud
Numbers that read well on paper can be awkward to say. Practise each headline number aloud until it comes out smoothly, with the baseline first and the outcome second. Notice whether you stumble over units or percentiles, and simplify where you do. A number said fluently sounds like something you know well, while a number said hesitantly sounds like something you looked up the night before, even when it is accurate.
Numbers reveal judgment about what mattered
The numbers you choose say something about you. A candidate who reports that a change cut p99 latency on the checkout path, and that this mattered because slow checkouts were abandoned, shows they understood what the business needed. A candidate who reports only that a refactor reduced code size shows a narrower view. Interviewers read the choice of metric as evidence of how you think about the purpose of engineering work.
A quick self test
Read the last sentence of each story in your bank. If it contains no number, no count and no concrete change, the story is not finished yet.
A number lets the interviewer write down something defensible, shows that you measured, and sets the scope of the work. Vague results are among the most common reasons strong stories score below the bar. Use two to four numbers per story, in the setup, actions and result.