Where we are
You have a skills matrix and a growth goal. The career conversation is where those connect to something the person actually wants, which is usually a level, a title, or a kind of work. Managers avoid this conversation for an understandable reason: they do not control promotions, so it feels dishonest to discuss one. That gets it backwards. You do not control the outcome, but you control almost everything the outcome depends on - which work they get, whether the evidence is documented, whether anyone in the levelling room has heard of them. Refusing to have the conversation does not protect anyone; it just means the person plans their career without you, and increasingly somewhere else.
Translate the level into things you can see 7 min live
Your organisation's ladder is probably written at a level of abstraction that makes it unusable in a Tuesday conversation. Your job is to translate it once, for your own function, into behaviours specific enough that the person can check themselves against them.
LiveThe running evidence file3 min▶
Promotion cases assembled in the two weeks before a levelling meeting are weak for a structural reason: they are built from memory, which means they are built from the last quarter and the most visible work. The fix is unglamorous and takes about four minutes a month.
- Keep a running note per person with dated examples mapped to the level criteria. When something happens, add a line. That is the whole system.
- Share it with them. This is the part managers resist and it is the part that changes behaviour: if the person can see which criteria have two examples and which have none, they will go looking for the missing evidence themselves.
- It fixes recency and it fixes your own memory bias - the sibling OKR course covers why a manager's recollection is a poor instrument, and a dated file is the cheap remedy.
- It also protects them if you leave. A person whose evidence lives only in their manager's head loses everything when that manager changes, which is one of the most common ways good people get stuck.
The promotion that was lost in a handover. A Cartwheel analyst had been told for a year she was close. Her manager left, and the new one arrived with no history, no file, and reasonably declined to advocate for someone he had known for six weeks. She was effectively reset to zero and left within the year. Nothing unfair was intended by anyone. The evidence simply did not exist anywhere except in a departed manager's memory.
How to say "not yet" and keep someone 5 min live
The difference between a "not yet" that motivates and one that starts a job search is entirely structural. Four components, in this order, and the order matters - leading with the gap makes everything after it sound like consolation.
LiveThe two hardest cases3 min▶
- The person who is close. Tempting to over-promise, because you want them to stay and you believe it will happen. Do not. Say "the evidence for X is there, Y is the gap, this project would close it, and if it lands I will put you forward in the spring cycle" - then be specific that you are putting them forward, not that they will get it. Promising an outcome you do not control converts a supporter into someone who feels lied to.
- The person who is not close and does not know it. The hardest conversation in this session, and delay makes it worse in both directions - they invest years in the wrong expectation, and your eventual honesty looks like a sudden reversal. Lead with what is true, be exact about the gap, and be honest about the size of it: "this is more than one project away, and I want to be straight with you about that rather than let you plan around next cycle." Then ask what they want to do with that information. Sometimes the answer is a different route entirely, and that is a good outcome reached early.
In both cases the failure mode is the same - vagueness offered as kindness - and it is the same failure the OKR course finds in feedback language generally: comfortable for the manager, expensive for the person.
Self-studyWhen the honest answer is "not here"2 min read▶
Sometimes the level or the work someone wants does not exist on your team, and no amount of development changes that. A small team may have no room above them for years; a specialism they want may not be part of your remit.
- Say it. "The role you want does not exist here and I do not expect it to for at least two years" is a hard sentence that people consistently report being grateful for, because it lets them decide with real information.
- Then help. Internal moves, a secondment, an introduction elsewhere in the organisation. A manager who helps someone leave well gets a reputation that makes the next hire much easier - and people talk.
- Do not hoard. Blocking an internal move to protect your own delivery is the single fastest way to lose someone entirely rather than lose them to another team.
- Expect the counterintuitive result: teams known for developing people out of themselves tend to attract better candidates than teams known for keeping everyone in place.
Translate a Cartwheel level and place someone on it ★ 12 min · everyone tries
Take the analyst to senior analyst jump and do the translation for your own function, then map one person's actual evidence against it.
Write each criterion in language that names an observable behaviour in your function's work.
For one person, list the dated evidence you actually have per criterion. Note the empty rows.
For each empty row, name a piece of real upcoming work that would fill it.
Check the access question: is that work available to them, or does it require someone to hand it over? If the latter, that someone is you.
| Criterion, translated | Evidence held | Work that would close it |
|---|---|---|
| Analysis changes team decisions, not one buyer's number | Two examples, both inside one category | Lead the analysis on the cross-team returns project |
| Given a question with no method, chooses and defends an approach | Strong - the autumn forecast, built from scratch | Already met |
| Work is reproducible by another analyst from the documentation | None. Nobody has ever tried to rerun her work | Hand the seasonal model to Dana to run once |
| Reviews a colleague's work and it measurably improves | Informal only, nothing recorded | Formal reviewer on Dana's category plan this quarter |
Script the harder "not yet" ★ 8 min · do your own
Write the four-part version for someone on your team who wants a promotion they will not get this cycle. Then read it aloud and check for the two failure words.
Write part one: the criteria they have genuinely met, named specifically. This has to be real, not warm-up.
Write part two: the gap, in the translated language, with no hedging.
Write part three: the work, and what you will do to make it available.
Write part four: an honest timeframe. If it is more than a year, say more than a year.
Read it back and delete every "hopefully" and every "should be fine" - those are the two words that convert a route into a promise you cannot keep.
The two responses to the same honesty. Cartwheel's team lead ran the four-part conversation with two analysts in the same week. One responded by taking the returns project and being promoted eleven months later. The other listened, said the timeframe did not work for her, and moved to a different function internally where the route was faster. Both conversations were successes, and the second one would have been read as a failure by most managers - which is why people avoid the honesty. The alternative was not keeping her; it was keeping her for another eighteen months while she quietly disengaged and then left anyway, without the internal move.
Try it yourself - this week ◐ 30-45 min total
- Translate one ladder level into observable behaviour for your function. Do it once and reuse it in every career conversation.
- Start a running evidence file for each person, with dated examples against criteria. Four minutes a month.
- Share one person's file with them and watch what they do about the empty rows.
- Run one four-part "not yet" conversation, and strike out "hopefully" before you say it.
- Identify anyone whose route genuinely does not exist on your team, and tell them - then help.
- Check whether the work that closes anyone's gap is actually available to them, or whether you are the person holding it.
Sources covered (cited)
Taught from ladder practice and the promotion-equity research. This session covers:
Three questions before you go 🎯 ◐ 90 seconds
1 · Managers avoid career conversations because they do not control promotions. Why is that backwards?
Refusing the conversation does not protect anyone. The person plans their career without you, and increasingly somewhere else. Promise the route and your part in it, never the outcome.
2 · What is the point of sharing the running evidence file with the person it is about?
It also survives a manager change, which is one of the most common ways good people get stuck - evidence that lives only in a departed manager's memory resets to zero when they leave.
3 · "You are doing great, keep going, hopefully next cycle." What has this told the person?
Four parts, in order: what is already true, the specific gap, the work that would close it, an honest timeframe. Vagueness offered as kindness is comfortable for the manager and expensive for the person.