You are the least reliable witness to your own conversations

You come out of the one-to-one and you think it went well. She nodded. Nobody cried. You got through the thing you had been carrying around since Monday, and the relief is so strong that it feels like evidence.

Three weeks later the work has not changed, and you find yourself explaining to your own boss that you did raise it, definitely, you had a whole conversation about it.

Your own rating is a moderate guide, nothing more

In 2014 two psychologists, Ethan Zell and Zlatan Krizan, pooled 22 meta-analyses comparing what people said about their own ability with how they actually performed: academic work, medical skills, language, sport, jobs. The average correlation across all of it was .29, with the individual results running from .09 to .63.

That is not nothing. People are not random about themselves. But .29 is the statistical equivalent of a rough steer. It is enough to know that you are probably better at planning than at confrontation. It is nowhere near enough to tell you what happened in a room for forty minutes on Tuesday.

The part of that finding you can use

The same paper found that self-ratings got closer to reality when the question was specific to a narrow domain and when the performance being judged was objective and familiar.

Which tells you exactly what sort of question to stop asking yourself. “Am I good at difficult conversations?” is the worst-shaped question available: broad, about character, and impossible to be wrong about. Nobody has ever answered it and learned anything.

Behaviour questions are the ones that work. After the next hard one, ask:

• How long was it before I said what the conversation was about?
• Did I describe a thing that happened, or a way that they are?
• Did I ask for their side before I had finished making my case?
• What did I say straight after the first time they pushed back?
• How much of the talking was mine? Roughly a third, half, most of it?
• Did anyone say who was doing what, by when, out loud?

Six answers you can actually get wrong. That is the whole point of them. This is also why Steady’s self-check asks about behaviours rather than confidence: a question you can be wrong about is the only kind that finds anything.

Other people know things, but not if you ask like that

The obvious fix is to ask your team. The obvious fix does not work, because “how am I doing?” from the person who writes your review is not a question, it is a small test of loyalty, and everyone passes it.

Lara Hogan’s questions for a first one-to-one get round this by being specific, practical, and asked long before anything is at stake. How do you like to receive feedback? What does a bad day look like for you, and how would I know you were having one? What did your last manager do that you want me to keep doing?

Ask those in week one and you have a record of how each person wants to be told things, written down before you ever have to tell them something they do not want to hear.

The things you would never think to check

Linda Hill spent years following new managers through their first year for Becoming the Boss. The pattern she found was not incompetence. It was a set of beliefs that felt obviously true and were not: that the title carries the authority, that managing means the team does what you say, that the job is the old job with more of it.

None of that shows up in a self-rating, because a self-rating can only search the part of the map you have already drawn. The gaps that matter are the ones you would not have thought to look for. Somebody, or something, has to point at them.

The honest counter-argument

Marcus Buckingham and Ashley Goodall have argued that other people are unreliable raters too: when I rate your “strategic thinking”, most of what the number measures is me. They are right, and it is worth taking seriously before you go and collect eleven opinions about yourself.

It does not rescue self-assessment. It points somewhere else. If your rating of you is moderate and their rating of you is noisy, then the trustworthy evidence is neither: it is the behaviour itself. How long you took. How often you hedged. Whether a next step was ever said out loud.

What to do this week

Before your next hard conversation, write down the three things you are going in with: what happened exactly, what you need to see instead, and the one question you will open with. Afterwards, answer the six questions above from memory, and note the time you started and the time you got to the point.

Do it four times. The fourth set will disagree with the first in at least one place, and that disagreement is the most useful thing you will learn about yourself as a manager this quarter.