How to ask for honest feedback

When you ask a person for honest feedback, you are asking them to accept a real cost on your behalf. If they say the true thing and you take it badly, they have damaged a relationship they value in exchange for nothing.

They have also watched people ask this question before. Most of those people wanted reassurance and said the word honest to make the reassurance feel earned. Your friend cannot tell which kind you are, and the safe guess is the common one.

So the polite answer is not a failure of courage. It is a correct reading of the risk, made in about a second, by somebody who likes you.

Ask people who have seen you in different settings. A colleague, somebody who knew you before this job, and one person who has seen you in a bad week will give you three genuinely different accounts.

Do not ask only the people who like you most. Warmth and honesty are not opposites, but somebody with a lot invested in your good opinion has the strongest reason to round the answer up.

Ask more than one person, and this is the part most people skip. A single account is one relationship described from one side. When three people who do not know each other reach for the same word, that word is about you rather than about any of them.

There is a real research finding underneath this. Connelly and Ones published a 2010 meta-analysis in Psychological Bulletin on ratings made by people who know the person being rated. Those ratings carry information self-ratings do not, and they predict some outcomes better than a person's own answers.

They also found that the gain comes from the number of raters and how well they know you. One acquaintance is weak evidence. Several people who know you in different contexts is strong evidence, and the strength comes from the independence.

Kruger and Dunning showed in 1999 that people least able at a task are the worst judges of their own ability at it. The skill you would need to assess the work is the same skill you are missing.

Their paper is about competence rather than personality, and it does not say everyone is deluded. What it establishes is narrower and enough: self-assessment fails in exactly the places you cannot see, and those are the places you were hoping to learn about.

Ask about something that is finished. A project still running is not a memory, it is a live position, and any comment on it moves the thing along. You will get a negotiating line dressed as an observation.

Two to six weeks after the fact is the useful window. Long enough that nobody is defending a decision, recent enough that the details are still there. Ask about last year and you get a summary somebody has already told twice.

Avoid the hour after anything went badly. The answer you get then is about the incident and about repairing the mood, and both of you will treat it as a conversation about blame.

A message is easier for both of you. They can think for an hour, delete a first attempt, and send the second one. You read it without your face in the room, which removes the test in move four.

Face to face has one real advantage, which is that you can ask a follow-up immediately. It also has a cost: they are watching you in real time, and they will adjust mid-sentence when something registers.

Send one question, not a list. A list reads as work, and work gets answered on the easiest item. Say when you want it by, and say that a short answer is fine.

No rush on this. Thinking about the launch review in March, what did I do in that room that made your job harder?

Here is the mechanics on a single example. The bad version is the one almost everybody sends first.

Before. Hey, do you have any feedback for me? I really want to hear it, be brutally honest.

Three things are wrong with it. There is no occasion, so they have to pick one, and picking one is the hardest part of the job you just handed them. There is no scope, so a small answer feels like a dodge and a large answer feels like an ambush. And brutally honest is a request for permission to be hurt, which raises the stakes rather than lowering them.

After. The version below fixes each one in order.

When I took over the client call in March and started answering for you, what was that like from where you were sitting? I think I did it twice.

The occasion is named, so no selection is required. The scope is one behaviour on one afternoon, so a two-line reply is a complete answer. And the admission is already on the table, so agreeing with it is not an accusation.

At some point somebody will describe something you are certain did not happen that way. You will want to explain the context, because the context is real and they did not have it.

Do not. You did not ask what happened, you asked what it was like for them, and on that question they cannot be wrong. Their experience of you is the thing you were trying to find out, and you just found it out.

There is a practical reason too. The moment you argue, you have converted a witness into an opponent, and opponents stop reporting. They will not tell you the next one, and the next one might have been the one that mattered.

If the facts genuinely matter, separate the two conversations by a week. Take the account today, check the record later, and never in the same breath.

That is not how I remembered it, which is probably the useful part. What did it look like from your side?