Writing Key Results
A Key Result is not a task. It is a measurable change someone can see in the world. These four sections cover what that means in practice: how to frame the result, how to choose the right indicators, how to handle missing baselines, and how to stay within what your team can actually move.
Who does what, by how much
A Key Result names a measurable change: who is doing what differently, and by how much. The test is simple. Imagine the Key Result is done. What is different for a real person, and how would you see it in a number? If finishing the work alone is enough to tick the box, you have written an activity, not a result.
Ask, then wait: "If this is done, what will be different for someone, and how would we notice?"
- Do not rewrite the Key Result for them. The question does the work.
- Sort each proposed Key Result into "result" or "activity" on cards, no debate during the sort.
- The disagreements that surface afterwards are the lesson.
Example: turning an activity into a result
Launch the new onboarding flow this quarter.
New users who finish onboarding in their first session rise from 41 to 65 percent.
The second version names who (new users), what changes (finishing onboarding in the first session), and by how much (41 to 65 percent). Launching the flow is only one way to get there. If the launch ships but the number does not move, the team has learned something. With the activity version, they would have called it done and learned nothing.
A result everyone can watch move is where focus, transparency, and alignment actually come from.
Leading and lagging indicators
Lagging indicators confirm the outcome at the end of the cycle: retention, repeat purchase, churn. Leading indicators move earlier and hint whether you are on track, because they capture the behaviour that tends to cause the lagging result. A Key Result set made up entirely of lagging indicators gives the team nothing to steer by mid-quarter. A set made up entirely of leading indicators risks optimising a proxy that never pays off. Aim for at least one of each, and label which is which.
For each Key Result, ask: "Will this number move in week three, or only at the end of the quarter?"
- If everything only moves at the end, the team is flying blind mid-quarter.
- Help them identify one early signal they can actually watch and act on.
- Name each indicator as leading or lagging so the intent is visible to anyone reading the OKR set.
Example: a steerable leading and lagging pair
Repeat purchase rate rises from 9 to 16 percent. (This is only visible at quarter end.)
Repeat purchase rate rises from 9 to 16 percent. Plus: first-time buyers who return to browse within 7 days rise from 22 to 35 percent. (The second KR is the early signal.)
By week four the team can see whether browse-return rate is moving. If not, they can adjust before the quarter ends. The lagging indicator alone would give them no such chance.
Transparency mid-quarter depends on having something that moves before the verdict lands.
Baselines and targets
A target without a baseline is a guess dressed up as a commitment. If the team does not know the starting number, the first Key Result is to find it: measure the baseline for X by a given date, then set a real target. Replace the placeholder once the number exists. Where uncertainty is high, a range or a floor beats a false precision figure that nobody believes.
When a team writes "increase X to 50 percent" and cannot say today's number, ask: "What is it today?"
- If they do not know, do not let the invented target stand. Make baseline discovery the first Key Result.
- A baseline-finding KR is honest and removes the theatre of invented numbers.
- Once they have the real number, the team replaces the placeholder with a target grounded in evidence.
Example: when you do not know the number yet
Raise trial-to-paid conversion to 25 percent. (From what? Nobody knows.)
Measure trial-to-paid baseline by week 2 and set a quarter target by week 3. Once measured: trial-to-paid rises from 12 to 20 percent.
The team commits to something real. The placeholder version gives a number that looks precise but is invented, which means any end-of-quarter grade is meaningless. Admitting you do not yet measure something is an act of transparency, not a weakness.
Transparency starts with admitting what you do not yet measure.
Stay within your influence
A Key Result should measure a behaviour the team can actually move. If the number depends entirely on another team's decision, it is an audit metric, not a commitment. This applies to internal teams too: if your team serves colleagues rather than external customers, the behaviour change you measure is in those colleagues. Where work genuinely spans teams, share the OKR rather than pretend you own it alone.
Ask: "Whose behaviour has to change for this Key Result to move, and can your team influence them?"
- If the answer is "another team, and no", either reshape the Key Result or propose a shared OKR.
- Name dependencies out loud rather than burying them in a footnote.
- For internal teams: ask who the internal customer is, and write the Key Result as a change in that person's behaviour, not as a description of your own work.
Example: an internal team's Key Result
Improve cross-functional collaboration this quarter.
Cross-functional initiatives that are unblocked within 5 days of escalation rise from 40 to 75 percent.
The second version names a concrete behaviour (unblocking within 5 days), a measurable range (40 to 75 percent), and implies who is affected. The team can act on it directly. "Improve collaboration" gives no signal and no grip.
Alignment means owning what you can move and being honest about what you share.