How to Run an OKR Retrospective on Goals, Not Just Work
How to run an OKR retrospective: four questions at the end of each cycle, a mid-quarter honesty checkpoint, and the courage to call a goal dead in week five.
To run an OKR retrospective, spend about half an hour at the end of each cycle on four questions about the goals, not only about the work, and hold a shorter honesty checkpoint around the middle of the quarter. Naming a dead goal early is part of the same practice.
Most teams hold some kind of review of the work. Far fewer review the goals themselves, and that is where the learning is. This article covers the end-of-cycle session, the mid-quarter checkpoint, and the permission to declare a goal dead that connects them.
How to run an OKR retrospective: the four questions
The guide's retrospective is four questions, asked at the end of each cycle.
| Question | What it surfaces |
|---|---|
| What did we set out to change, and did it change? | Whether the goals described outcomes and whether those outcomes happened |
| Where were our estimates wrong, and in which direction? | A pattern of over or under ambition |
| What did we learn about how long things actually take here? | The real pace of change in your business |
| What should the next set look like as a result? | A direct input to the next planning session |
Notice what is missing: any question about who did well or badly. The session is about the goals and the estimates, which is why it can be honest. It is also why it depends on the separation from pay and reviews. Where scores feed assessments, answers to the second question become defensive.
Why retrospect on the goals and not only the work
A review of the work asks whether the team delivered what it planned. That is useful and incomplete, because a team can deliver an enormous amount and change nothing. A review of the goals asks whether the thing you wanted to be different is different, and what that tells you.
The guide is clear that this is where institutional memory is created. A year of written goals, honest scores and short retrospectives becomes a record of what you set out to do, what happened and how good your estimates were. It takes half an hour, and skipping it is why companies run OKRs for three years and are no better at setting them than they were in the first quarter.
The scores that feed this session need to be honest, as described in how to score OKRs. Inflated scores make the first question meaningless.
Running the end-of-cycle session
Keep the session short and in a fixed order so the evidence comes before the opinions.
- Read each key result with its baseline, target and score.
- Take the four questions in turn, writing the answers down as you go.
- For the second question, look for a direction: did you consistently aim too high, or too low?
- For the fourth question, produce notes the next planning session will actually use, such as which kinds of goal to avoid or how much to attempt.
- File the notes where the next cycle's author will find them.
The last step matters more than it looks. The record is only valuable if the next author reads it. The questions are cheap, and the answers are the point.
Treat a low score on an aspirational goal as information, not failure. The guide notes that such a score is frequently a good outcome. And if everything scored well, start with the ambition question, because an all-green quarter is a warning and not a triumph.
The mid-quarter honesty checkpoint
The guide adds a second, shorter session around the halfway mark. It is slightly longer than a weekly check and it has one question only: given what we now know, are these still the right goals, and is anything about to fail that we are not saying out loud?
Its justification is practical. Most teams already know the answer by then and are waiting for permission to say it. The checkpoint is that permission, and it recovers half a quarter that would otherwise be spent maintaining an agreed fiction.
Make the session a scheduled event, not something that happens only when someone is brave. The checklist for the quarter asks for four things at this point.
- You have asked whether these are still the right goals given what you now know.
- Anything clearly dead has been declared dead, in writing, with the reason.
- Effort released from a dead goal has been redirected deliberately.
- Anything at risk has been said out loud rather than carried quietly.
Naming a dead goal early
Sometimes a key result becomes impossible or irrelevant. The market moved, an assumption was wrong, or something more important arrived. The guide observes that this happens every quarter in every company, so the failure is not that it happens.
The failure is spending many more weeks pretending, because abandoning something feels like admitting error. The guide's remedy is to make it explicitly acceptable to declare a goal dead mid-quarter, state why in writing, and redirect the effort.
A team that can kill a goal in week five is faster than one that carries a corpse to week thirteen. The written reason has a second use, since it is worth more in the retrospective than a completed goal would have been: it records what you learned and how early you were willing to say so.
How to make it safe to say a goal is dead
People will only say it if saying it is safe. Three conditions help.
- Scores are not tied to pay, reviews or promotion, so declaring a goal dead costs nobody anything.
- The decision is written down with its reason, so it is a record and not a retreat.
- The freed effort goes somewhere deliberate, so the declaration leads to a plan.
The signal usually appears first in the weekly check, as confidence that has dropped and stayed dropped. A team that has been through how to run a weekly OKR check in will have a place to say it before the checkpoint arrives.
What to have ready before you start
A retrospective is only as good as the record it draws on. Before the session, gather the material the guide has already asked you to keep.
- The key results with their baselines and targets, unchanged since the start.
- The score for each, taken without adjustment.
- The written reasons for any goal declared dead during the quarter.
- The notes from the mid-quarter checkpoint, including what was said out loud.
If any of these are missing, note it. A retrospective that cannot answer the first question because the baselines were never written is itself a finding, and the right response is to fix the process, not to invent the answer from memory.
Keeping the notes useful
The fourth question produces the notes that matter, and notes only help if someone reads them. Write them as short statements about the company, not about individuals, and phrase them so that the next author can act on them.
Good notes record a direction ("we tend to aim too high") and a consequence for the next set ("so label more goals aspirational and expect to fall short on some"). They avoid adjectives about people. Over time the notes become the institutional memory the guide describes, a record that tells new joiners not only the plan but how the company thinks.
What the retrospective feeds
The retrospective is not an end. It is the input to the next set of goals. The fourth question asks what the next set should look like, and the answer should shape three things the guide has already covered.
- How many goals to attempt, in light of what you learned about pace. See how many OKRs a startup should have.
- How ambitious to be, and which goals to label committed or aspirational.
- Which baselines you still lack, since a goal to find a number may be the right first key result of the next cycle.
Used this way, the retrospective turns a series of quarters into a learning loop. Without it, you are writing the same quarter again with different dates.
When the retrospective keeps repeating the same lessons
If the same answer appears quarter after quarter, such as always aiming too high, treat it as a finding and change the next set accordingly. The guide's account of the long-term benefit rests on this: you learn whether you systematically overestimate, which bets have worked and how long change really takes.
If the lessons are repeated and nothing changes, look at the nine mistakes in OKR mistakes founders make. A retrospective that is skipped, or whose notes are never read, is a mistake in itself.
Frequently asked questions
How long should the retrospective take?
The guide puts the end-of-cycle session at about half an hour. The mid-quarter checkpoint is slightly longer than a weekly check, and it covers one question only.
Is this the same as a sprint retrospective?
The guide's retrospective is about the goals: what you set out to change and whether it changed. A review of the work alone misses that.
When should we declare a goal dead?
When it has become impossible or irrelevant. Say so as soon as you know, state the reason in writing and redirect the effort, rather than carrying it to the end.
Who should be in the room?
The guide does not specify attendees. Include the owners of the key results, since they hold the facts the four questions need.
The full method, with the checklists, is in The OKR Playbook, a free 32-page playbook.