.png)
Overview
Every OKR book says the same thing: don't tie goals to performance reviews or pay.
Most companies do it anyway. If that's you, this article is not going to tell you to stop. It's going to show you how to do it without breaking the goal system you've spent two years getting people to use.
Outcomes-based performance management means rating people on the goals they delivered, weighted for difficulty and context, as one input alongside how the work was done. It is not completion percentage in, rating out. That distinction is the whole article.
We'll cover:
Why the OKR orthodoxy says no
Why companies ignore it, and why that's not entirely wrong
The one prerequisite: named owners
Eight rules for rating outcomes fairly
Where finance fits in
Why the books say don't
Andy Grove built the goal system that became OKRs at Intel. John Doerr took it to Google and wrote Measure What Matters. Both are clear: keep OKRs separate from compensation and formal reviews.
Their reasoning is sound. The moment a goal decides your rating, three things happen:
Sandbagging. People set targets they know they can hit.
Safe ambition. Stretch goals disappear because a stretch goal is now a risk to your bonus.
Gaming. Progress gets reported to the number, not to reality.
A goal system is supposed to push people toward ambitious, uncertain outcomes. Attaching a rating to it quietly turns it into a to-do list of things everyone already knew they could do.
Why everyone does it anyway
We have yet to meet a company that keeps the two apart in practice.
"What did you deliver this year?" is the most defensible question a manager can ask in a review.
It's concrete. It's hard to argue with. And when a manager has to justify a rating to HR, to the employee, or to a calibration meeting, delivered outcomes are the evidence they reach for first.
So goals influence ratings whether the policy says so or not. The difference is whether it happens formally, with visible rules, or informally, where nobody can see how the sausage gets made. Informal is worse. It has all the sandbagging risk and none of the transparency.
The real problem isn't linking outcomes to reviews. It's linking them badly.
How does this fail? Completion percentage goes in, rating comes out. 87% on your goals, that's a 4 out of 5, next.
NO! BAD! That version deserves everything Grove and Doerr said about it. The rules below are how you avoid it.
The prerequisite: every goal has a name on it
Before any of the eight rules work, you need one thing in place. Every goal that can be rated must have a single named owner.
A goal owned by "the marketing team" is owned by nobody. Nobody updates it, nobody gets asked when it stalls, and at review time, nobody can be fairly rated on it because five people were all half-responsible.
Ownership doesn't mean doing all the work. It means being the one person accountable for the outcome and for keeping the goal current. Contributors do the work; the owner carries it. If your goal system only has team-level goals, you can't do outcomes-based performance management. You can only do outcomes-based guessing.
Get this right first. Then the rest becomes possible.
Eight rules for rating outcomes fairly
1. Split committed goals from aspirational ones
Google's own OKR practice distinguishes between committed goals (expected to be hit, full stop) and aspirational goals (a 70% hit rate is a good result).
Rate delivery on committed goals. Treat aspirational goals as context: what did they attempt, how far did they get, what did they learn. Never turn a stretch goal into a score. The day you do, nobody sets another one.
2. Rate the judgement, not the number
70% on a hard goal beats 100% on an easy one. Everyone knows this and almost nobody's rating process reflects it.
The manager's job is to assess difficulty, context, and what was in the person's control. A progress bar can't do that. If your review form pulls in a completion percentage and asks the manager to confirm it, you've replaced judgement with arithmetic. Ask the manager to rate the outcome and use the number as one input to that rating.
3. Set goals before the period, rate them after
No retrofitting. Goals written in review week to match what already happened aren't goals, they're a retrospective with extra steps and a waste of time.
This sounds obvious but it gets violated constantly, usually because goal-setting slipped in January and by March nobody wanted to admit there were no goals. Fix the calendar, not the review.
4. Make outcomes one input, not the whole rating
Somewhere around half the weight is reasonable. The rest goes to how the work was done: behaviours, competencies, values, whatever your organisation has decided matters.
Pure outcomes rewards the lucky and punishes the people who took on the hardest problems. It also tells your team that how they treat colleagues is irrelevant as long as the number lands. Most companies don't mean that. Their weighting says it anyway.
If you're building a rating scale for this, our guide to performance rating scales covers how to structure one.
5. Re-scoping is fine. Deleting is not.
Priorities change. A goal set in January can be legitimately irrelevant by June. Changing it, with a visible record of what changed and when, is healthy. It's what a live goal system should do.
Quietly removing missed goals in November is something else entirely. The moment people notice it's possible, every goal in the system loses credibility, because everyone knows the ones that survived to review are the ones that were going to look good.
Keep the history. A goal that was re-scoped mid-year with the reason logged is evidence of good management. A goal that vanished is evidence of the opposite.
6. Calibrate across managers
Goal difficulty varies wildly by team. One manager sets aggressive goals and rates a near-miss as strong performance. Another sets comfortable goals and rates every hit as exceptional. Without calibration, the second team gets the bonuses.
Calibration is where the easy-goal teams get caught. HR or leadership reviews rating distributions across managers before results are released, and asks the awkward questions. It's also the step most small companies skip, because it takes a meeting. It's worth the meeting.
7. Keep development goals out entirely
Development goals ("get better at presenting to senior stakeholders") have no business in the document that decides someone's rating. Nobody writes a truthful development goal in that document. They write a safe one.
Keep growth plans in a separate space with a separate purpose: coaching, not compliance. It's why we built Growth Plans as a permanent, always-on plan that never gets pulled into a review round. Development conversations only work when the stakes are low.
8. No compensation formula
Ratings should inform pay decisions through judgement. They should not determine pay through a spreadsheet.
The moment there's a formula (rating 4 = 6% increase), everyone optimises for the formula. Managers inflate to protect their people. Employees argue the rating instead of the work. And leadership loses the ability to make sensible exceptions without it looking like favouritism.
This is also why we've long recommended separating the review conversation from the compensation conversation into two meetings. Same principle, applied to the calendar.
Where finance fits in
A note from a former CFO: outcomes-based performance management is the only version finance actually believes in.
Behaviour-only reviews read to a finance leader as opinion. Outcomes are the thing that connects the people budget to results, which is the question finance is always asking. If your Head of People needs budget for the performance process, this is the framing that gets it: we rate on delivered outcomes, we weight it sensibly, we calibrate across teams, and we don't let a formula make pay decisions.
That's a system a CFO will fund. A pile of spreadsheets with completion percentages is not.
If you only do three things
Most companies can't fix all eight at once. If you're heading into a year-end cycle in the next couple of months, start here:
Put a named owner on every goal. Without it, nothing else is fair.
Rate the judgement, not the number. Change the review question from "confirm completion" to "rate this outcome given its difficulty."
Calibrate before you release results. One meeting, all managers, rating distributions on the wall.
Do those three this cycle and add the rest next year.
Frequently asked questions
Should OKRs be tied to performance reviews?
The OKR orthodoxy says no, because linking goals to ratings encourages sandbagging and safe targets. In practice, most companies do it anyway, and doing it informally is worse than doing it openly. OKRs can feed performance reviews if committed goals are separated from aspirational ones, managers rate outcomes with judgement rather than by completion percentage, and ratings are calibrated across teams.
What percentage of a performance rating should be based on goals?
Around half is a reasonable starting point. Weighting outcomes at 100% rewards luck and punishes people working on the hardest problems; weighting them near zero tells finance the review is opinion. The remainder should reflect how the work was done, through behaviours, competencies or values.
What is the difference between committed and aspirational OKRs?
Committed OKRs are expected to be fully achieved and can fairly be rated on delivery. Aspirational (stretch) OKRs are deliberately ambitious, where reaching around 70% is a good result. Aspirational goals should be discussed as context in a review but never scored, or people will stop setting them.
TeamMaven's Goals and Reviews modules were built around exactly these rules: every goal has a single named owner, goals can be pulled into a review as a rateable element with the manager's judgement on top, every change to a goal or rating is logged, and HR calibration runs across the whole round before anything reaches employees. Live in two weeks, EU-hosted, and priced for a scaling company rather than an enterprise.
