How well do you know your team?

You can't grade your read on a person the way you'd grade a trivia answer; there's no key. What you can check is whether the picture you're working from is current and specific or old and general, and that check takes ten questions and about ten minutes.

Pick one direct report. The most interesting choice is the person you've managed longest: the picture should be deepest there, and it's usually the one running furthest behind. Write the answers down; "I could answer that if I had to" is exactly the feeling this test exists to check.

The ten questions

What are they working on this week, and which part of it do they care about most?
What's their biggest frustration right now?
What do they want next — more scope, a different kind of work, a quieter year — and did they tell you that, or did you infer it?
What did they do in the first month of the current review period that belongs in the review?
When did they last surprise you, and how?
When did you last change your mind about them, and what changed it?
Which of your beliefs about them are you most confident in, and what's the newest evidence you have for it?
What are they better at than you?
Which working relationship on the team costs them the most energy right now?
If they resigned next month, would you be surprised, and what's the newest thing you know about how they feel about being here?

Checking your answers

Three different checks, depending on the question.

Ask (1, 2, 3, 9). Take them into the next 1:1 as questions, not as theories to confirm: "what's eating most of your energy lately?" beats "I've been assuming X, right?" Compare what comes back with what you wrote.

Look it up (4). Their commits, docs, demos, your own calendar. This one has a correct answer, and the lookup is the point: if what you find surprises you, that's recency at work.

Check the dates (5, 6, 7, 10). These can't be verified against anything external, so check the timestamps instead. Not what you answered, but when the evidence behind it is from.

Question 8 has no check at all. It's on the list because a blank there usually means the picture was built around your own strengths, and the person's edge sits somewhere you don't often look.

Reading the misses

No score; the misses carry more information than a count would. Three patterns are worth noticing.

Everything you wrote is from the last three weeks. Your picture is running on recency. That's normal, well documented, and the whole subject of the evidence page.

Your most confident belief has the oldest evidence. Certainty that hasn't been re-confirmed in months is a habit. Take that read into your next 1:1 as a question.

The forward-looking answers were inferences (3, 10). You've been managing a model of the person, and the model stopped getting inputs. This one is also the easiest to fix, because both answers are a question away.

Keeping the picture current

Nothing here needs software. A doc per person and five minutes after each 1:1 covers most of it; Lara Hogan recommends exactly that, and her first-1:1 questions are the fastest way to fill the blanks this test found. Rands' essay on 1:1s is good on the half of this that happens in the room. A monthly skim of each doc, oldest lines first, keeps the dates honest. The template has the fields, and the how-to has the method. Retake this in a quarter; the second run tests the habit rather than the memory. And if the misses trace less to habit than to headcount, there's a page for that.

Full disclosure: I make Continuum, a Mac app that keeps the dates for you: each belief carries its confidence and its latest evidence, signals fade from view unless you notice them again, and the app's front page is question 7 asked continuously: which beliefs have gone longest without fresh evidence. It scores no one, including you. Free for up to three people (Pro is a one-time purchase if you'd rather not subscribe), and nothing ever leaves your Mac. The doc per person above is free and works; the app is for having the dates checked without remembering to check them.