OKRs vs. KPIs: What's the Difference?

OKRs vs KPIs comes down to one question: are you keeping something healthy, or trying to change it? A key performance indicator (KPI) is an ongoing measure of how well something that already runs is doing, such as how fast a support team answers. An OKR (objectives and key results) is a time-boxed push to change something, set for a quarter or a year and then graded. John Doerr’s team, which promotes the method, makes the same split: OKRs measure change, KPIs measure health.¹

You rarely have to choose. The two work best as a pair, because a good key result usually moves a KPI that matters, and the other KPIs keep watch so the push does not break something else. If you are new to goals of any kind, the habits that help in every format are in the evidence-based way to set goals.

Prefer to watch? The video covers the same ground as this article.

VideoOKRs vs KPIs: What's the Difference?WiserHours

OKRs vs KPIs side by side

OKRs and KPIs differ in purpose, time span and what counts as success. A KPI runs continuously and should stay in a healthy range. An OKR lasts a fixed cycle and aims beyond the current level; Google’s guide scores each key result between 0 and 1 and expects an average of roughly 0.6 to 0.7 across all OKRs.²

Wikipedia’s entry defines a KPI as a type of performance measurement used to judge how well an organization, project or process is meeting its objectives, with quality, efficiency and customer satisfaction as typical examples.³ Google’s re:Work guide pairs an ambitious, slightly uncomfortable objective with key results that can be measured and scored as a number.²

Keeping the two apart matters because their jobs pull in opposite directions. A health measure works best when nobody is straining to push it, so a change in it tells you something real. A push works best when it is narrow, temporary and a bit uncomfortable. Ask one number to do both and it does neither well: the team either coasts on a target it already hits or strains forever at a number that was only meant to be watched.

A practical first step is to label every number your team reviews as “watch” or “push”. Most will be “watch”. The handful marked “push” are your candidate key results for the next cycle.

Where OKRs and KPIs come from

OKRs trace to Andy Grove, who introduced them at Intel in the 1970s and described the approach in his 1983 book High Output Management. John Doerr learned the method at Intel, introduced it to Google around 1999 and set it out in his 2018 book Measure What Matters.⁵ KPIs have no single inventor: they grew out of ordinary performance measurement.

Grove’s version, as Google’s guide summarizes it, rests on two questions: where do I want to go, and how will I pace myself to see if I am getting there? The first gives the objective, the second the key results.² Doerr’s team adds that Grove’s “iMBOs” (Intel management by objectives) differed from older management by objectives in being quarterly rather than annual and separate from compensation.¹

One widely used framework for choosing KPIs came from Robert Kaplan and David Norton in a 1992 Harvard Business Review article on the balanced scorecard.³ Their opening point was that a company’s measurement system strongly shapes how managers and employees behave, and that traditional financial measures can give misleading signals about improvement and innovation.⁶ That warning applies to both tools.

Little is known about OKRs specifically. When the Brazilian researchers Roberto Silva and Gleison Santos mapped the academic writing on OKRs in 2024, they counted 47 studies on its use, judged the method’s theory poorly documented and found little that examines it closely. The paper their map describes as weighing the similarities, differences and pros and cons of OKRs and KPIs was an opinion article.⁷ Most of what follows therefore leans on wider goal-setting research, which is far stronger, and on company practice, which is labeled as such. If a tool vendor says switching to OKRs will lift your results, ask which study shows it.

Can a KPI be a key result?

Yes, and often it should be. In a 2018 piece in MIT Sloan Management Review, Donald and Charles Sull note that some companies use metrics or KPIs in place of key results, and argue that the label matters less than translating each goal into concrete measures of progress.⁴ What changes is the number’s job: as a KPI it is watched, as a key result it is pushed.

The pairing makes sense for a reason goal research spells out. Edwin Locke and Gary Latham, who developed goal-setting theory, concluded in their 2002 review that goals paired with feedback on progress are more effective than goals alone.⁸ A KPI dashboard is that feedback. An objective with no number behind it leaves people unable to tell whether they are on course; a number with no objective behind it tells them where they are but not where to go.

1. The KPI dashboard: ongoing health measures you watch regularly; most stay in their normal range
2. The OKR: one KPI that needs to change becomes a key result for a fixed cycle, next to the objective it serves
3. The guardrail: another KPI the push must not damage, watched while the OKR runs
4. Back to the dashboard: once the new level holds, the number goes back to being watched, not pushed

Not every KPI belongs in an OKR. Google’s guide warns against “business-as-usual” OKRs that describe what a team would achieve without changing anything. It accepts that an objective such as keeping customer satisfaction above a set level may stay the same quarter after quarter, but says the key results under it should change to push the team further.²

Promote a KPI to a key result only when it has drifted out of range or needs a new level, and keep one or two other KPIs as guardrails that the push must not damage.

A worked example: a customer support team

This example is illustrative, not drawn from a real company. A support team’s KPI dashboard shows four numbers: first-response time, customer satisfaction after each ticket, the backlog of open tickets and the share of tickets reopened. Three sit in their usual range. One does not: customers keep coming back about the same three problems, password resets, billing questions and a confusing export feature.

The team does not write an OKR for the whole dashboard. It writes one for the number that is off:

• Objective: customers can fix our three most common problems without contacting us twice.
• Key result 1: repeat contacts about those three problems fall from about 1 in 4 to about 1 in 10.
• Key result 2: a help article for each problem, rated useful by most readers who vote.
• Key result 3: every agent trained on the new fix steps by week four.
• Guardrails: first-response time and satisfaction stay in their normal range.

Key result 1 is a KPI on loan. When the quarter ends and the new level holds, it goes back on the dashboard with a new normal range, and next quarter’s OKR can address something else. The guardrails matter because pushing one number invites shortcuts elsewhere, such as marking tickets solved before the customer says the fix worked; the mechanics of that are in how team metrics backfire under Goodhart’s law.

Myth

OKRs replace KPIs.

Fact

OKRs sit on top of KPIs: the dashboard keeps running while one or two numbers get a push for a cycle.

As a team lead, write the objective about the customer or the result, borrow the KPI as the measure, and name the guardrails out loud in the planning meeting.

A worked example: a personal goal

The same split works for one person, with smaller numbers. This example is also illustrative. Someone who wants to move into a team-lead role keeps two personal KPIs they check each Friday: hours of focused work in the week, and evenings free of work messages. Both keep their week healthy.

The change goes into an OKR for the next six months:

• Objective: be ready to apply for a team-lead role by the summer.
• Key result 1: lead two small projects from kickoff to delivery.
• Key result 2: ask three colleagues for written feedback on how I run meetings.
• Key result 3: finish one course on giving feedback.

The two KPIs now act as guardrails. If the projects eat every evening, the push is costing more than it should. Locke and Latham’s review warns that performance suffers when two goals demand actions that cannot both be done, so it pays to decide in advance which one gives way.⁸ For writing and grading the objective itself, see the guide to personal OKRs.

Should you use OKRs or KPIs?

Use KPIs alone for steady work you only need to keep healthy; add OKRs when something must change; and in most teams, use both, with OKRs drawing their key results from the KPI dashboard. Sull and Sull describe companies that set goals in different ways and conclude that, at least for building ambition, no single approach suits every company.⁴

A payroll team, a compliance function or a mature service with satisfied users may need nothing more than a good dashboard and a regular review. A team launching a product, fixing a KPI that has been off for months, or learning a new kind of work benefits from an objective that says what is changing and why.

Ask whether this number would stay where you need it if nobody pushed it for a quarter. If yes, watch it. If no, it may deserve a key result.

How many OKRs should a team have?

Google’s re:Work guide suggests three to five objectives with about three key results each, and warns that more can over-extend teams and diffuse their effort.² Those figures are company practice, not tested limits, and they describe Google’s organizational OKRs; a single team usually needs fewer.

Research on several goals at once explains the caution. The 2009 critique “Goals Gone Wild”, by Lisa Ordóñez and colleagues, reports research in which people with multiple goals tended to concentrate on one, and a lab study in which participants given difficult quantity and quality goals sacrificed quality to hit the quantity target.⁹ A long list of key results does not get all of them done.

That is also why turning every KPI into an OKR backfires. A team with twelve “key results” that are its whole dashboard has no priority at all, and the uncomfortable objective the method relies on disappears into routine reporting.

Are OKRs tied to performance reviews?

In the original design, no. Google’s re:Work guide says plainly that OKRs are not synonymous with employee evaluations.² Doerr’s team describes OKRs as divorced from compensation.¹ Sull and Sull report that Google deliberately decouples goal attainment from reviews and pay, while the brewer AB InBev ties bonuses tightly to its targets and builds ambition in other ways.⁴

Most organizations do link goals to pay. Sull and Sull cite a 2013 survey by the consultancy Mercer of 1,056 organizations in 53 countries, mostly for-profit and weighted toward large employers, in which most said performance on goals influenced appraisals, promotions or bonuses.⁴ That sits awkwardly with OKRs that are meant to fall short, and the reviewnarrative review: A review in which an expert summarizes a body of research in prose, choosing which studies to discuss. It can be a good map of a field, but without a stated search and pooling method it is easier for selection and opinion to shape the conclusions than in a systematic review.Full entry in the glossary of goal research by Locke and Latham shows where it bites.

The study

Limited evidence

Locke and Latham on paying for very hard goals, 2002

When a goal is very difficult, paying people only if they reach it can hurt performance: once they see the reward slipping away, their personal goal and their belief that they can do it drop, and so does their work. The drop did not appear when the goal was moderately difficult, or when people with a difficult goal were paid for each unit of performance rather than for hitting the target.⁸

The finding comes from the theory’s own developers summarizing their field, not from a test of OKRs, so read it as a pointer, not proof. It fits what Sull and Sull describe: requiring 100 percent of a target for a bonus encourages people to set conservative targets they are sure to reach.⁴

If pay has to reflect performance, base it on KPIs and on contribution, and keep stretch OKRs out of the bonus formula. Grade OKRs in the open and treat a low grade as information for next cycle, which is how Google’s guide frames it.²

Before your next planning cycle

• Label every number the team reviews as watch or push
• Pick one to three numbers to push; leave the rest as KPIs
• Write an objective that says why each push matters
• Name one or two guardrail KPIs the push must not damage
• Keep OKR grades out of the bonus formula
• Return each pushed number to the dashboard once its new level holds

The bottom line

A KPI tells you whether something that already runs is healthy; an OKR names what you will change this cycle and how you will measure it. Most teams need both, with key results borrowed from the dashboard and the rest of the dashboard watching as guardrails. Keep the OKRs few, and keep them out of the bonus formula.