Designing a leveling rubric that survives growth
Leveling rubrics rot at scale. Here's the design pattern — common dimensions, function-specific examples, and the calibration discipline — that keeps the bar…
- Five dimensions, applied to every role: scope, complexity, impact, influence, autonomy.
- Examples per dimension per level — function-specific, written by people at that level.
- Refresh the examples annually; the structure rarely changes.
- Test the rubric: read three real packets through it; if it doesn't produce a clear answer, fix the rubric.
A good leveling rubric is a tool managers actually use, not a poster on the wall. The trick is making it specific enough to differentiate adjacent levels and general enough to apply across functions.
The five dimensions
| Dimension | Asks | Anti-pattern |
|---|---|---|
| Scope | What's the size of what you own? | Tenure dressed as scope |
| Complexity | How hard are the problems you solve? | Counting projects, not problem class |
| Impact | How much does it matter to the business? | Activity confused with outcome |
| Influence | How much do you change others' work? | Loudness confused with leadership |
| Autonomy | How much guidance do you need? | Free-for-all confused with senior |
How many levels?
5-7 levels for IC, 4-5 for management, mapped together. More than 8 levels at <500 people means you're paying for theatre — engineers will optimise for the next promo, not the next problem. Fewer than 4 means you can't differentiate the new-grad from the principal.
Function-specific examples
- Core rubric is shared (5 dimensions, level descriptors).
- Per function (eng, design, sales, ops): real examples of what each level looks like.
- Written by current job holders at that level, not by HR alone.
- Reviewed annually for staleness; updated for new scope/tech.
Stress-testing the rubric
- Take 3 real packets from the last cycle.
- Re-evaluate using the new rubric, blind to the prior outcome.
- If the answer matches the prior outcome, rubric agrees with current practice.
- If it diverges, decide: was prior decision wrong, or is rubric wrong?
- If two reviewers disagree using the rubric, the rubric is the problem, not the reviewers.
Senior IC and management are parallel paths to the same level, not stepping stones. The rubric should make both equally achievable and equally compensated. If your principal engineers earn less than your directors, your ladder is single in practice.
- The promotion process: nominations, packets, committee, and decision
- Career Ladders That Don’t Trap People
- Calibration Sessions Run Well: The Hidden Operating Layer of Performance Management
- Locke & Latham's goal-setting theory — the science behind (and against) OKRs
- The manager effectiveness scorecard: what to measure and how often
- Performance management 101: what it is, what it isn't, and where to start
Read next
All playbooksPromotions get political when the process is informal. Here's the end-to-end design — nomination, packet, committee, decision, communication — that holds up…
How to design IC and management ladders that give people a real path, hold a consistent bar, and avoid becoming a filing system for politics.
Calibration is the meeting that decides whose ratings are real. Done well, it removes manager bias and produces defensible decisions on pay, promotion, and…