From "measure fairly" to "write the goal"
Last session you learned to tell activity from outcome and to run a metric through the fair-metric checklist. Tonight you turn that judgment into the format your company actually writes goals in: the OKR - Objective and Key Results. The craft is small but unforgiving. Most bad OKRs fail in one of two ways: the Objective is a vague wish, or the "Key Results" are really a to-do list wearing a number. We will fix both on Cartwheel, then set the scoring scale that tells you at quarter-end how you did. One rule carries over and never leaves: an OKR score is for steering and learning. It is decoupled from pay and ratings - the moment a score decides a bonus, people sandbag the goal and stop being honest.
The Objective inspires; the Key Results measure 7 min live
John Doerr's formula is deliberately plain: "I will [Objective] as measured by [Key Results]." The Objective is qualitative, memorable, and time-boxed - it says where you are going and why it matters. The Key Results are the numbers that prove you arrived. Get the split right and the rest of OKR practice is easy.
LiveThe "done" test - outputs in disguise3 min▶
The single most common OKR mistake is dressing a task up as a Key Result. Use one test:
- Can it be checked off as "done"? "Ship the new checkout", "run five campaigns", "launch the app" - these are outputs. You either did them or you did not, and doing them proves nothing changed.
- Does it name a measurable change? "Checkout completion 61% → 72%", "conversion 2.1% → 2.8%" - these are outcomes. The team could ship every feature on the roadmap and still miss them, which is exactly why they are worth grading.
Shipping is not succeeding. A Cartwheel squad once "hit" its quarter by shipping a redesigned checkout - then discovered completion had not moved a point. The feature was the plan, not the goal. Key Results measure the result the feature was supposed to buy. Keep the roadmap as your task list; keep the OKR as the outcome it should produce.
Committed vs stretch, and the 0-1 scale 6 min live
Not every OKR carries the same promise. Google re:Work separates two kinds: committed OKRs you fully expect to hit (target score 1.0), and aspirational or stretch OKRs you set knowing ~0.7 is a strong result. Grading both on the same 0-1 scale only works if everyone knows which kind they signed up for.
Self-studyThe ~0.7 norm is Google's - adapt it, and keep it away from pay2 min read▶
Two things about scoring get misused, so hold them carefully:
- ~0.7 is a Google convention, not a law. It works because Google explicitly decouples OKR scores from pay, so a "0.7" carries no penalty. If your organisation is newer to OKRs, start with more committed goals and introduce stretch gradually. Pick a norm, state it out loud, and keep it consistent so scores are comparable.
- The score is a conversation-starter, never a verdict. A 0.5 with a good story (the market moved, a dependency slipped) can be a better quarter than a 1.0 on a goal that was too easy. Read the score with its trend and context, exactly as you learned in b1.
And the linchpin again: because OKRs are meant to be ambitious, wiring the score to a bonus guarantees sandbagging. Keep scoring on the steering track. Pay is decided elsewhere, on a fuller picture.
Self-studyFewer, bolder OKRs - 3 to 5 objectives, not 151 min read▶
OKRs are a focusing tool, so the count matters. Doerr's guidance: 3 to 5 objectives per team per quarter, each with 2 to 4 Key Results. If everything is a priority, nothing is - a team with fifteen objectives has simply relabelled its to-do list. When you cannot cut, ask which OKR you would defend if you could only keep three. Those are your real goals; the rest is ordinary work that does not need an OKR to get done.
Rewrite a weak Cartwheel OKR ★ 12 min · everyone tries
A Cartwheel team lead brings a first draft: a vague Objective and a to-do list masquerading as Key Results. Walk it through the anatomy and the "done" test, then rewrite it.
The draft. Objective: "Do well in Q3." Key Results: "ship the new checkout", "run 5 marketing campaigns". Score it against the anatomy from Part 1.
Diagnose the Objective: not inspiring, not time-specific in any meaningful way, gives no direction. "Do well" cannot be aimed at.
Diagnose the Key Results with the "done" test: both can be checked off as done. "Ship checkout" and "run 5 campaigns" are outputs - they prove effort, not results.
Rewrite the Objective so it inspires and points somewhere: "Make Cartwheel's checkout the smoothest step of the journey."
Rewrite each Key Result as a measurable outcome with a from → to number: conversion rate 2.1% → 2.8%, checkout completion 61% → 72%. The shipping and the campaigns move down to the roadmap as the how.
| Part | Before (weak) | After (fair + measurable) |
|---|---|---|
| Objective | "Do well in Q3" | "Make checkout the smoothest step of the Cartwheel journey" |
| KR1 | "Ship the new checkout" (output) | Conversion rate 2.1% → 2.8% |
| KR2 | "Run 5 campaigns" (output) | Checkout completion 61% → 72% |
| Passes "done" test? | ✗ both can be ticked off | ✓ both name a measurable change |
Label Cartwheel OKRs: committed vs stretch ★ 8 min · do your own
Setting the commitment level is a promise, so it has to be explicit. Take a mixed set of Cartwheel Key Results and label each as committed (expect 1.0) or stretch (expect ~0.7), with the score you would call a good quarter.
For each Key Result, ask: is this something we must deliver (committed), or a bold reach we set to pull ourselves further (stretch)?
Write the expected score next to each so no one is surprised at quarter-end - committed targets 1.0, stretch aims for ~0.7.
Sanity-check: if every OKR is committed, you are playing it safe; if every one is stretch, nothing is guaranteed. Aim for a mix.
| Cartwheel Key Result | Type | Good-quarter score |
|---|---|---|
| Keep site uptime ≥ 99.9% | Committed | 1.0 - a floor, not a reach |
| Ship PDPA-compliant consent flow by quarter-end | Committed | 1.0 - a must-do |
| Lift conversion rate 2.1% → 2.8% | Stretch | ~0.7 (reach ≈ 2.6%) is strong |
| Grow repeat-purchase rate 28% → 38% | Stretch | ~0.7 is a genuine win |
Name the type before the quarter, not after. The unfairness sneaks in when a team treats a stretch goal as committed only once it is missed - suddenly a healthy 0.7 reads as failure. Cartwheel's fix is a one-line label on every OKR at the start. It costs nothing and removes an entire category of quarter-end argument.
Try it yourself - this week ◐ 30-45 min total
- Take one OKR your team wrote and run every Key Result through the "done" test. Rewrite any that can be checked off as done into a measurable outcome.
- Check your Objective: is it qualitative, inspiring, and time-boxed? Rewrite it if it reads like a task or a wish.
- Label each of your current OKRs committed or stretch, and write the expected score next to it.
- Count your objectives. If you have more than five, decide which three you would defend and demote the rest to ordinary work.
- Optional: read Google re:Work "Set goals with OKRs" and Doerr's "Measure What Matters" ch. on scoring before b3.
Sources covered (cited)
Taught from the OKR canon on goal anatomy and scoring. This session covers:
Three questions before you go 🎯 ◐ 90 seconds
1 · "Ship the new checkout" is a weak Key Result mainly because...
The "done" test: if a Key Result can simply be ticked off, it is an output. Rewrite it as the measurable change it should cause, such as checkout completion 61% → 72%.
2 · On a stretch (aspirational) OKR, what score is generally considered a strong result?
Around 0.7 is the Google stretch-goal sweet spot (a norm to adapt, not a law). A 1.0 on a stretch goal usually means it was not ambitious enough. Committed OKRs, by contrast, target 1.0.
3 · Why must OKR scores stay decoupled from pay?
OKRs are meant to be ambitious. Wire the score to a bonus and people set safe goals to protect the number. Keep scoring on the steering track; decide pay elsewhere, on a fuller picture.