Calibration has landed one of your engineers a lower rating than you argued for. How do you deliver it to them?

The rating you deliver has to be yours

There is one move that ruins this conversation and almost everybody reaches for it, because it feels kind. It is some version of "I put you forward for higher and the committee disagreed". You have now told the engineer three things you did not intend: that the outcome was arbitrary, that their manager cannot affect their career, and that there is nobody they can usefully talk to about it. You have bought yourself a moment of shared grievance at the cost of every future conversation about their performance, because from here on any feedback you give can be dismissed as something you do not really mean.

The position that works is that you deliver the rating as the organisation's assessment, which you are now accountable for explaining, and you explain it on the evidence rather than on the process. If they ask what you argued, you can be honest that you made the case for more and that having heard the comparison across the group you understand where it landed — provided that is true. What you cannot do is describe the outcome as a mistake and then ask them to accept it.

Where you genuinely believe the outcome is wrong and not merely unwelcome, the place to say so was in the calibration room, and the thing to do afterwards is fix the input for next cycle rather than relitigating this one with the person affected. Telling an engineer their rating is unjust while being unable to change it is the cruellest available option: you have given them the grievance and no remedy.

Do the preparation the conversation actually needs

Before the meeting, be able to answer three questions without hesitating, because the engineer will ask all three and a vague answer converts disappointment into distrust.

The first is what the rating rests on. Not a summary judgement, but two or three specific pieces of work and the specific expectation they did or did not meet, from whatever written level definition your organisation uses. "The migration you led was solid; the two design proposals you owned both needed me to resolve the cross-team disagreement, and at this level that resolution is expected to be yours" is a sentence someone can act on. "Overall impact was slightly below the level" is not.

The second is what a different outcome would have required. Concretely, in terms of work that existed or could have existed this year.

The third is what happens next, and here honesty about mechanism matters. Say what you control — the assignments they get, the visibility of their work, the case you write — and say what you do not, which is the comparison against everyone else at that level.

Phrasing that fails	Phrasing that holds
"The committee went a different way on yours."	"This is my assessment and I will take you through what it is based on."
"Impact was a bit light this year."	"Two things carried it: the migration landed, and the two designs both needed me to close them out."
"Keep doing what you are doing and it will come."	"The gap is owning a cross-team decision end to end. The billing integration next quarter is that, and it is yours."
"I know, it is a bit unfair, the process is what it is."	"I made the case for higher. Having seen the whole group, I can explain where the line fell, and here is what changes it."
"I'll push hard for you next cycle."	"I control the assignment and the written case. I do not control the comparison, so I am not going to promise you an outcome."

A surprise is a management failure, and you should say so first

If the rating is news to the engineer, the substantive problem is not the rating. It is that they spent a year believing something untrue about how their work was seen, and every week you did not correct that was a week they could have spent differently. Name that yourself, early in the conversation, before they get there: you should have told them in the autumn that the design work was not landing at the level, you did not, and that is on you.

Managers avoid this because it looks like undermining the rating. It does the opposite. It is the only thing that makes the rating credible, because it explains why the news is arriving now, and it is the strongest signal you can give that the next twelve months will be different. It also protects you from the version of this conversation where the engineer discovers the concern was in your notes for six months, which converts a difficult review into a broken relationship.

The mechanism that stops it recurring is a running record. Write down what you would say about someone's performance every month or two, and once a quarter say it to them. A manager who does that never has this conversation cold, and a manager who does not will have it again next year with somebody else.

What you offer, and what you must refuse

The engineer will want reassurance, and the reassurance they want is a promise about the next cycle. Do not give it. A promise of a rating or a promotion you do not control is the single most reliable way to lose someone twice: once when it does not materialise, and again in the way they describe you to everyone else afterwards.

What you can offer is specific and within your gift. One assignment that produces exactly the evidence that was missing, agreed in writing with what good looks like. A commitment to a short check-in on that assignment on a stated cadence, so nothing accumulates silently. A named point mid-year at which you will tell them honestly whether the case is on track, including if the answer is no.

Then deal with the retention risk plainly rather than by hoping. If they are strong and disappointed, they are now more likely to look elsewhere, and pretending otherwise does not help. The useful question is whether the constraint is the rating or something behind it — scope, a project they hated, a peer who takes their work. Ask, and be willing to hear that the honest answer is that this organisation cannot give them what they want, in which case helping them move internally is a better outcome than a resignation in three months.

One case deserves separate handling: the engineer whose year was ruined by a cancelled project or a reorganised roadmap. They are usually right that it was not their fault, and the rating may still be a fair description of the evidence available, which is an uncomfortable combination. Do not resolve it by pretending the evidence exists. Resolve it by owning the allocation decision that put them there, saying that the next assignment is chosen so this cannot repeat, and taking the argument about how your organisation rates people whose projects are cancelled into the calibration process rather than into the room with the person who lost out.