Where we are
You have decided the manager is the system and that coaching is the mechanism. Coaching against what? A manager cannot develop someone toward a standard that does not exist, and "be more senior" is not a standard. This session builds the reference document. Two things to hold as we go. First, a ladder is a description, not an aspiration - it should describe how your organisation actually works, otherwise nobody can use it. Second, the ladder is where pay equity is either built in or quietly broken, because every level carries a band, and every vague level definition is a place where two people doing the same work end up at different levels for reasons nobody wrote down.
Levels are scope of impact 7 min live
Published ladders across many industries converge on the same underlying idea: what changes as you go up is not how hard you work or how long you have been there, but how far the effects of your judgment reach, and how much ambiguity you can absorb before someone else has to.
LiveThe four things a level definition must contain3 min▶
A usable level definition is not a paragraph of adjectives. It answers four questions, each in language a manager and a report could both check against reality:
- Scope - whose work is affected when this person makes a judgment? Their own, the team's, several teams', the function's.
- Autonomy - how much ambiguity do they absorb? Given a task, given a problem, given a goal, or expected to find the problem themselves.
- Craft - what technical or professional standard is expected, expressed as observable output rather than a label.
- Effect on others - what happens to the people around them. At junior levels, nothing is required. Above a certain point, lifting others stops being optional and becomes the level.
If a definition has all four and still cannot settle an argument, it is because the words are not observable. "Demonstrates strategic thinking" fails; "identifies which of several competing problems the team should work on, and can explain why the others can wait" can be checked.
Self-studyThree failure modes that make a ladder unusable2 min read▶
- The adjective ladder. Every level says the same thing with a stronger adverb - "consistently", "highly", "exceptionally". Nobody can tell where they are, so levelling defaults to the manager's advocacy and the loudest case wins.
- The wish-list ladder. Written as who the organisation wishes it had, not how it actually operates. If nobody at your company has ever done what your top level describes, the level is decoration and the ladder loses credibility all the way down.
- The too-many-rungs ladder. Eleven levels between junior and principal means the difference between adjacent rungs is invisible, so promotion becomes an annual administrative event rather than a real change in scope. Fewer, wider levels are easier to defend and harder to game.
A fourth, which is really a3 meeting a4: a ladder with no dual track. If every rung above a certain height requires managing people, you have decided that senior craft is worth less than management, whatever your values statement says.
A shared standard, not a rank order 6 min live
A ladder on paper drifts within a year unless levelling is calibrated. But calibration goes wrong in a specific and predictable way, and the failure is worth naming precisely because it looks like rigour.
LiveThe policy you have to write down3 min▶
Five rules, and they are the same five the sibling OKR course uses for rating calibration, because the failure modes are identical:
- The level definition is fixed before anyone is discussed, so the standard cannot move to fit a conclusion.
- Evidence per claim. "She operates at the next level" is not a claim until it names work and its reach.
- No quota. If a pay budget genuinely constrains promotions, that is a budgeting decision made in a separate, clearly-named step - not smuggled into the levelling conversation.
- Examine outliers in both directions - the manager who levels everyone high and the one who levels everyone low.
- Write down every change and its reason. A level that moves without a recorded reason is a new bias, not a correction.
For the mechanics of rater bias underneath all of this - why one manager's judgment differs so much from another's - the sibling course's session on beating the bias does that work properly, and it is worth reading before you chair one of these rooms.
Write one Cartwheel level properly ★ 12 min · think it through
Take the jump everyone argues about - analyst to senior analyst - and write it with all four components in observable language. Then test it the only way that matters.
Write the scope: whose work is affected when this person exercises judgment.
Write the autonomy: what they are handed. A task, a problem, or a goal.
Write the craft as observable output, not a label.
Write the effect on others, and be explicit about whether it is required at this level or not.
Test it: give the definition to two managers with the same three people and see whether they level them the same way. If not, the words are not observable yet.
| Component | The adjective version | Observable version |
|---|---|---|
| Scope | "Broader impact" | "Their analysis changes what the merchandising team decides, not only what one buyer does" |
| Autonomy | "Works independently" | "Given a business question with no defined method, chooses the approach and defends it" |
| Craft | "Strong analytical skills" | "Produces analysis another analyst can rerun from the documentation, with stated assumptions and limits" |
| Effect on others | "Is a role model" | "Reviews at least one colleague's analysis per cycle and their work measurably improves as a result" |
| Tenure | "3-5 years' experience" | Not a criterion. It appears nowhere in the definition |
Stress-test your existing ladder ★ 8 min · your org
Four fast tests. Any ladder that fails two of them is not being used for real decisions, whatever the intranet says.
The two-manager test. Do two managers level the same person identically from the document alone?
The self-assessment test. Can a report read it and locate themselves without asking their manager?
The external-hire test. Can you level a candidate from outside using it? Tenure-based ladders fail this immediately.
The top-rung test. Has anyone here ever actually operated at your highest level? If not, it is decoration.
The promotion that revealed the ladder was fiction. Two Cartwheel analysts with near-identical work were levelled differently in the same cycle. Asked to justify it from the ladder, both managers pointed to the same sentence - "demonstrates broader impact" - and read it in opposite directions, each entirely sincerely. The pay gap that resulted was not created by bias in the ordinary sense; it was created by a document that could not settle an argument. That is the practical case for observable language: it is not pedantry, it is the thing standing between your ladder and an equity problem you will find eighteen months later in a pay audit.
Try it yourself - this week ◐ 30-45 min total
- Run the two-manager test on the level in your organisation with the most disputed promotions. Note where the two managers diverge and which word caused it.
- Search your ladder for the word "consistently" and for any years-of-experience range. Rewrite one instance into observable scope.
- Count your rungs. If adjacent levels cannot be told apart in a sentence, plan a merge.
- Check whether your highest individual-contributor level has ever been awarded. If not, decide whether to fix it or delete it.
- Write the five calibration rules into your levelling process, especially the no-quota one, and name where budget decisions happen instead.
Sources covered (cited)
Taught from published ladder practice and the calibration literature. This session covers:
Three questions before you go 🎯 ◐ 90 seconds
1 · What should the spine of a career ladder be?
Published ladders converge on scope, not service. A tenure spine rewards staying, promotes people into scope they have never held, gives managers nothing to coach toward, and makes external candidates impossible to level.
2 · "Demonstrates strategic thinking" is a poor level criterion because...
The test for any level definition is whether two managers could independently agree who is at it. Observable language ("identifies which competing problem the team should work on and can explain why the others can wait") settles arguments; adjectives hand the decision to whoever advocates hardest.
3 · Leadership tells the levelling room that only 15% may sit at the senior level. What have they done?
Calibration exists to level the standard, not to allocate a budget. Hand the room a quota and the output becomes a rank order in which the most confident advocate wins. If a budget genuinely constrains promotions, make that a separate, clearly-named decision.