One person on the team is said to be the problem, most of what you have is second-hand, and there is a date in the prompt — so the answer is a sequence, not a verdict.
When two juniors tell you a colleague is making them miserable, you have a report, not a finding. It is real information about the two juniors and thin information about the third person. The mistake senior candidates make is to jump straight to the hard conversation — which is only one step inside the intervention.
The whole intervention is a week. Establish what is actually known and name what is second-hand. Hear the person before you form the plan. Brief your own chain before you involve anyone else's. Protect the team without pretending you have concluded anything. Then close: what happens next, by when, and how you will both know it worked.
Pick a scenario, then place one move per day. Tap a move, then tap the next open day — or drag it. Each placement tells you what that move actually reveals or costs.
Five steps. The order is the content — most of the marks in a situational interview are for the sequence, not the outcome.
Write two columns before you decide anything. Second-hand accounts are not worthless; they are simply a different kind of thing, and they support a different set of decisions.
what i have first-hand | what i was told (second-hand)
--------------------------------+------------------------------
3 missed standups (i was there) | "makes us feel small" x2 juniors
14 review comments, read myself | "everyone has noticed" x1 relay
defect catch rate: highest | "she is checked out" x1 relay
decision this supports : open a conversation; gather specifics
decision it does NOT : pull the packet, change the rota,
state a verdict, open an HR file
Usually: to hear the person, and to make sure they hear it from you rather than about you. Name out loud what must not happen in it — a decision the facts do not yet support. Saying "I am not deciding anything today" is a leadership move, not a hedge.
The person concerned → the affected teammates → your own manager → HR or a peer manager. That order holds for ordinary performance and behaviour matters. It inverts the moment an allegation touches harassment, discrimination or safety: there, you notify first, you do not investigate, and you tell the reporter the limits of confidentiality before they rely on them.
Change the interface, not the person: rotate a reviewer, split a pairing, take a review yourself, rewrite the norm for everybody. Protection buys time and costs almost nothing — as long as the person hears it from you the same day.
close = the change + the date + the signal
"one named change: no verdict without a suggested fix.
we look again on the 14th.
the signal is that reviewees stop asking someone else to
translate your comments before they reply."
A week without a close is a week the person spends guessing, and the team spends assuming nothing happened.
| Situation | What the sequence gives you — and what it costs |
|---|---|
| A second-hand report about one person's behaviour | Converts worry into evidence and protects both sides. Costs two or three days you may feel you do not have. |
| A strong performer whose impact is harming teammates | Lets you keep the performance and fix the impact. Costs political capital, because someone will want a removal. |
| A conflict spanning two managers | Two managers running one intervention instead of two running two. Costs a day and requires you to share control. |
| Any allegation of harassment, discrimination or a safety risk | This sequence does not apply. Notify per policy without delay and hand the finding to people qualified to make it. You lose control of the timeline — correctly. |
| Performance that has not moved after a real, documented conversation | Move to the formal process. This framework is the thing that comes before one, never a substitute for it. |
Limitation: it assumes you have the authority to change the work, and that nobody is in immediate danger. If either is untrue, escalate on day one.
"Two of your juniors say your most senior engineer makes them feel small in code review. She has also gone quiet in standup. Her promotion packet goes to calibration on Thursday. What do you do?"
Start by naming the evidence: nothing first-hand, two second-hand accounts, one observable fact (three missed standups). Monday, go back to the juniors — not for permission, for specifics: dates, exact comments, and what they want to be different. That turns worry into something quotable without naming them. Tuesday, hear Dana, with the meeting's purpose stated in the first sentence and an explicit "I am not deciding anything today." You may discover, as often happens, that the withdrawal and the sharpness have one cause and it is not malice.
Wednesday, brief your manager — early, so it is information rather than rescue, and so calibration on Thursday holds no surprises. Also Wednesday: re-route the two juniors' reviews for a fortnight and tell Dana yourself the same day, so protection is not mistaken for a sentence. Thursday, the packet goes in with a note Dana has read. Friday, close: one named change, a check-in on the 14th, and the signal you will both watch — the juniors stop rehearsing before they submit. Say what would flip you to HR (a pattern, a protected characteristic, retaliation against the two who spoke), and say it as a threshold, not a mood.
You have heard nothing first-hand. Which sentence belongs in the first meeting with the person?
A team member tells you, asking you to keep it strictly between you, that a colleague has repeatedly commented on her accent and her age. What do you do first?