Somewhere this quarter a team wrote down "maintain 99.9 per cent uptime" and called it an OKR. It is not one. It is a KPI with a target attached, and the difference is the reason their quarterly review will feel pointless.
The two are not competing frameworks. They answer different questions and most companies need both. The trouble starts when one is used to do the other one's job.
What is a KPI
A key performance indicator is a number you watch continuously because it tells you whether something important is healthy. Monthly recurring revenue. Attrition by team. Average time to hire. Defect escape rate. Collection days.
KPIs have three properties worth naming. They persist, so the same metric matters this year and next. They have a threshold rather than an ambition, meaning there is a level below which you have a problem. And they are diagnostic, so when one moves you go looking for why.
A useful test: if the metric would still be worth reporting to the board in three years, it is a KPI.
What is an OKR
Objectives and key results is a goal-setting method that came out of Intel under Andy Grove and spread through Google. It has two parts. The objective is a short qualitative statement of what you want to be true by the end of the period. The key results are three to five numbers that prove you got there.
OKRs are time-boxed, usually quarterly. They are about change rather than maintenance. And they are meant to be uncomfortable, because a goal you were always going to hit tells you nothing you did not already know.
A KPI is the dashboard on your car. An OKR is the decision to drive somewhere different. Watching the dashboard harder does not change the destination.
OKR vs KPI: the difference in one table
| KPI | OKR | |
|---|---|---|
| Purpose | Monitor whether something stays healthy | Change something within a set period |
| Time frame | Ongoing, no end date | Usually one quarter, sometimes a year |
| Target | A threshold you must not fall below | A stretch you may not reach |
| Number of them | As many as the business genuinely needs | Three objectives at most per team |
| Failure means | Something is wrong, investigate now | The goal was ambitious, learn and reset |
| Typical owner | The function that runs the process | A named person for each key result |
| Ties to pay | Sometimes, carefully | Better not. See below |
The mistake almost everyone makes
The common failure is writing your existing KPIs as key results and calling the list an OKR set. It looks like the framework and produces none of the benefit, because nothing in it represents a decision to do something differently.
Here is the shape of it. Objective: "operational excellence". Key results: maintain uptime at 99.9 per cent, keep support response under four hours, hold churn below two per cent. Every one of those is a KPI. The team can hit all three by doing exactly what they did last quarter, which means the quarter had no goal.
The corrected version names a change. Objective: "support stops being the reason customers leave". Key results: reduce first-response time from four hours to forty minutes, cut repeat tickets on the same issue from thirty per cent to ten, raise post-resolution satisfaction from 3.4 to 4.2. Those are movements, from a stated start to a stated finish, and they might not happen.
The second failure is the reverse: turning genuine OKRs into permanent metrics. If the same objective appears three quarters running, either it was never achievable in a quarter or it has quietly become a KPI and should be moved.
When to use which
- Use a KPI when the thing must stay within a range and someone should be alerted when it does not
- Use an OKR when you want a specific number to move and you are willing to reprioritise work to move it
- Use a KPI for the machine that already runs: payroll accuracy, offer acceptance rate, statutory filing on time
- Use an OKR for the thing you keep saying you will fix and never get to
- Use neither for work that is simply going to happen. A roadmap item does not need a framework wrapped around it
How they work together
The practical arrangement is that KPIs generate OKRs. You watch the health metrics, one of them looks bad or plateaued, and that becomes the subject of next quarter's objective. When the OKR succeeds, the KPI settles at a new level and goes back to being monitored.
That relationship also gives you the guardrail. Every OKR set should list the KPIs that must not degrade while you chase it. A team that halves time to hire and doubles first-year attrition did not succeed. Naming the protected metrics up front prevents the argument afterwards.
How to write an OKR that is not a task list
- Write the objective as the change you want. One sentence, no numbers, memorable enough to repeat without reading it. If it could be a job description, it is not an objective
- Add three to five key results that measure the outcome. Each is a number moving from a stated start to a stated finish
- Check that each key result could fail. One you are already on track to hit is a forecast rather than a goal
- Name one owner per key result. One person, not a team, because shared ownership produces shared silence
- Cut until three objectives remain. A team with seven objectives has priorities, not a priority
The reliable tell for a task disguised as a key result is the verb. Launch, implement, complete, roll out and deliver all describe activity. Reduce, increase, move from X to Y describe outcomes. "Launch the new onboarding portal" is a task. "Cut time to first productive week from 21 days to 10" is a key result, and it might be achieved without the portal.
How to choose a KPI worth tracking
Most KPI dashboards are too long to be useful. The test for keeping a metric is whether anyone would do something differently if it moved. If a number can drop twenty per cent and nobody acts, it is reporting rather than a KPI.
- It has an owner who can actually influence it
- It has a defined threshold, agreed before the number goes bad
- It is reported on a fixed cadence to the same people
- It is a rate or a ratio more often than a raw count, because counts move with headcount
- It has a paired metric that stops it being gamed. Time to hire alone encourages rushed hiring, so pair it with quality of hire or first-year attrition
OKR vs KPI vs MBO and the balanced scorecard
OKRs did not appear from nowhere. Management by objectives, from Peter Drucker in the 1950s, established the idea of cascading agreed goals down an organisation. Its weakness in practice was an annual cycle tied tightly to compensation, which pushed everyone to negotiate soft targets.
The balanced scorecard, from Kaplan and Norton, is a KPI framework rather than a goal framework. It insists you measure across financial, customer, internal process and learning dimensions so no single number dominates. It pairs well with OKRs and is often what companies actually need when they say their KPIs are too narrow.
OKRs kept the goal cascade from MBO, shortened the cycle to a quarter, made the targets deliberately ambitious, and cut the link to pay. That last change is the one most companies quietly undo.
Should OKRs be tied to performance reviews and pay
Generally no, and the reason is mechanical rather than philosophical. OKRs only work if people set targets they might miss. The moment missing a target reduces someone's bonus, they will set targets they can hit, and the framework becomes an expensive way of writing down work already planned.
The usual objection is that goals nobody is accountable for get ignored. That is fair, and the answer is that OKRs should be visible, reviewed and discussed, which is a stronger form of accountability than a line in an appraisal form. Use KPIs and demonstrated judgement in the review conversation. Keep the ambitious quarterly targets out of the pay calculation.
If your organisation cannot separate the two, be honest about it and set OKRs at a level people can commit to. A committed OKR system that everyone trusts beats an aspirational one everyone games.
Where OKRs go wrong in practice
- Too many. Fifteen objectives across a team of eight means nothing was prioritised
- Cascaded mechanically, so a company objective is copied verbatim into nine team plans and owned by nobody
- Set in week one, reviewed in week thirteen, and forgotten in between. The weekly check is the habit that makes it work
- Graded like an exam, so a 0.7 becomes a failure rather than the intended outcome of an ambitious target
- Written by leadership for teams rather than with them, which produces compliance and no commitment
- Kept in a document nobody opens. If the OKRs are not in the tool people use daily, they are decoration
Frequently asked questions
What is the difference between OKR and KPI?
A KPI is an ongoing measure of health that you monitor against a threshold, such as churn or uptime. An OKR is a time-boxed goal for changing something, consisting of one ambitious objective and three to five measurable key results. KPIs describe the current state; OKRs describe the intended change.
Can a KPI be a key result?
Yes, when it is expressed as a movement rather than a level. "Churn below two per cent" is a KPI restated. "Reduce churn from 3.1 per cent to 2.0 per cent this quarter" is a key result, because it names a start, a finish and a deadline.
What are some good OKR examples?
Objective: new joiners become productive without needing to ask around. Key results: cut time to first shipped work from 21 days to 10, raise the 30-day onboarding satisfaction score from 3.2 to 4.3, and reduce setup tickets per joiner from 6 to 2. Each is a number moving between two stated points.
What are some good KPI examples for HR?
Attrition by manager and tenure band, offer acceptance rate, time to fill by role family, first-year attrition, payroll accuracy, statutory filings completed on time, internal mobility rate, and share of employees receiving recognition in a quarter. Each has an owner, a threshold and a fixed reporting cadence.
How many OKRs should a team have?
Three objectives at most, each with three to five key results. Beyond that the framework stops forcing the prioritisation that makes it valuable. If everything is an objective, the quarter has no objective.
Should OKRs be linked to performance reviews?
Preferably not. OKRs depend on people setting targets they might miss, and linking them to ratings or bonuses pushes everyone toward safe targets. Use KPIs and judgement in reviews, and keep ambitious quarterly goals separate from compensation decisions.
The short version. Keep a small number of KPIs that tell you whether the business is well. Pick the one or two that are not, and write OKRs to move them. Review both on a schedule, keep the ambitious ones away from the bonus calculation, and delete anything nobody has acted on in two quarters.
