The last mile of a fair system
Across five sessions you built the machine: measure what matters, tell OKRs from KPIs, decompose the metric tree to controllable levers, design against bias and gaming, and keep AI a helper. This session is the last mile - the part that decides whether the machine is trusted or feared. A fair metric checked once a year in a tense review is not a fair system; a rough metric discussed weekly with candour and safety often is. Rhythm and culture do most of the work. And the linchpin of the whole track holds here too: an OKR score is for steering and learning, never a direct input to pay - the instant it drives compensation, the honesty you need for any of this leaks away.
Cadence, maturity, and a crawl-walk-run rollout 7 min live
Monitoring has a natural rhythm: weekly 1:1s catch problems while they are small, monthly business reviews check the trend, quarterly OKR retros ask whether the goals were even the right ones. But the right frequency for a given person on a given task is not set by their rank - it is set by what Andy Grove called task-relevant maturity. New to the task? Check in often. Seasoned on it? Get out of the way. The same senior leader may need close support on an unfamiliar problem and near-total autonomy on a familiar one.
LiveThe shift from the annual review to continuous check-ins3 min▶
The annual review is the single worst place to do any of this - a whole year compressed into one high-stakes conversation, drowning in recency bias, arriving far too late to help. The industry moved for a reason:
- Adobe scrapped annual rankings for a lightweight "Check-in" in 2012, moving to frequent, forward-looking conversations.
- Deloitte (Buckingham & Goodall, HBR 2015, "Reinventing Performance Management") redesigned around frequent check-ins after concluding its old process consumed enormous time for little value.
- Gallup reports that employees who get frequent, meaningful feedback are far more engaged than those reviewed once a year.
The guardrails that keep measurement fair 6 min live
Culture is not a soft add-on to a measurement system - it is the thing that decides whether the numbers you get are true. Amy Edmondson's research on psychological safety is the sharpest version of this: on safe teams, bad news travels fast and early, so you can act on it; under surveillance and fear, bad news hides until it is a crisis. The four guardrails below are what a leader protects.
Self-studySeparating the three purposes at the org level2 min read▶
The manager track separates steering, development, and evaluation for one team. At the org level, the same separation is a design decision only leadership can make - and it usually shows up in structure:
- Steering lives in weekly 1:1s and monthly reviews - fast, cheap, forgiving of rough numbers.
- Development lives in ongoing coaching and growth conversations - forward-looking, never punitive.
- Evaluation (rating, pay, promotion) is rare, backward-looking, and deliberately fed by human judgment and calibration - not by raw OKR scores.
When an organisation lets the same quarterly OKR number flow straight into the compensation model, it has quietly merged all three - and every steering conversation for the rest of the year becomes a negotiation instead of the truth.
Match cadence to maturity, not to rank ★ 10 min · think it through
Grove's insight is uncomfortable for a hierarchy: the right monitoring frequency depends on how experienced the person is on this specific task, not on their seniority. Walk it through your own team.
Pick three people and one current task each. For each, judge their task-relevant maturity: new, developing, or seasoned on that task.
Set the check-in frequency from the maturity, not the title: frequent and detailed for new, lighter for developing, autonomous with a monthly trend-check for seasoned.
Notice the trap: reflexively giving your most senior person the least oversight on a task that is brand-new to them - that is rank overriding maturity, and it is where surprises come from.
| Cadence | Question it answers | Purpose |
|---|---|---|
| Weekly 1:1 | Is the work on track? What is blocked? | steering + development |
| Monthly review | Is the trend going the right way? | steering |
| Quarterly OKR retro | Were these even the right goals? | learning |
When NOT to measure ★ 8 min · your org
The most mature move a measurement leader makes is sometimes to not measure. A number that will be gamed, that no one controls, or that no one will ever act on is worse than no number at all - it costs trust and produces false confidence. Test your dashboards against three disqualifiers.
Will it be gamed? If the cheapest way to move the number does not require doing the real work, and it is tied to reward, expect Goodhart. Either pair it or drop it.
Is it uncontrollable? If the person or team cannot move it, grading them on it measures the weather, not their work. Watch it as shared context instead.
Will anyone act on it? If no decision will ever change based on the number, it is a vanity metric - it costs attention and monitoring effort and returns nothing.
Fewer, truer numbers. Teams instinctively add metrics to look rigorous, then drown in dashboards nobody reads. A leader earns trust by cutting the number of things measured down to the few that are controllable, hard to game, and actually acted on - and by being honest that some important things (culture, creativity, long-run loyalty) are watched and discussed, not scored. That honesty is the capstone of everything this track taught.
Try it yourself - this week ◐ 30-45 min total
- Draft your organisation's four-step fair-measurement rollout: baseline goals - OKRs decoupled from pay - fair metrics + cadence - continuous check-ins. Name where you are today and the next step.
- For each step, write the one condition that must be trusted before you climb to the next.
- Name one thing you will stop measuring people on - because it is gamed, uncontrollable, or nobody acts on it.
- Audit your check-in cadence: is any frequency set by rank when it should be set by task-relevant maturity?
- Optional: read Buckingham & Goodall, "Reinventing Performance Management" (HBR 2015), and Edmondson on psychological safety.
Sources covered (cited)
Taught from the cadence, continuous-PM, and culture canon. This session covers:
Three questions before you go 🎯 ◐ 90 seconds
1 · According to Grove, what should set the frequency of a check-in?
Task-relevant maturity, not seniority. A senior leader may need close support on an unfamiliar task and near-total autonomy on a familiar one. The rhythm follows the person and the task, not the title.
2 · Why does psychological safety matter for measurement?
Edmondson's research: safety makes people surface problems while they are still fixable. Surveillance and fear make the truth hide, so your numbers look fine right up until they do not.
3 · When should you decide NOT to measure something?
A number that will be gamed, that no one controls, or that no one acts on is worse than none - it costs trust and creates false confidence. Some things are watched and discussed, not scored.