OKR vs. KPI: Which Numbers to Hold Steady and Which to Move

Every key result scored 1.0, and nothing changed

Picture a customer support team's quarterly OKRs. Objective: deliver great support. Key result one: keep customer satisfaction above 90%. Key result two: answer every ticket within four hours. Key result three: keep the backlog under 200. At the end of the quarter, every key result scores 1.0. The team celebrates. Nothing about how it works is different from the quarter before.

Google's internal OKR guidance has a name for this. A goal written around “what the team believes it can achieve without changing anything they're currently doing” is a business-as-usual OKR, and the guidance lists it as a classic trap ². Those three numbers were never OKRs. They were KPIs with a deadline attached.

The direct answer: a KPI (key performance indicator) is an ongoing measure of how a part of the business is performing, watched against a normal range or a standard. An OKR (objectives and key results) is a time-boxed commitment to change something: an objective that sets the direction, and a few measurable key results that show, without argument, whether you got there. A goal is the plain-language destination either one serves.

The thesis of this page: a KPI is a promise to keep something healthy, and an OKR is a commitment to change something. If nothing has to be done differently to hit the number, it is a KPI, whatever the template calls it. Most of the confusion between the two, and most of the wasted OKR cycles, come from ignoring that one question.

If nothing has to change to hit it, it is a KPI wearing a deadline.

Three words, three jobs

John Doerr, who learned the system from Andy Grove at Intel ⁶, gives the shortest version: “The objectives are what you want to have accomplished. The key results are how I'm going to get that done. Objectives. Key results. What and how” ¹. KPIs need no such pairing. They are the gauges you already watch. For the textbook definitions, see the Lexicon entries on OKRs and key performance indicators.

The same number can sit in either column. Order accuracy is a KPI while you are protecting it and a key result when you are trying to lift it. What changes is the job the number is doing.

Where OKRs came from, and what the sources support

In his 2018 TED talk, Doerr says that in 1975 he was a computer engineer working under Andy Grove, and that “Andy invented a system called ‘Objectives and Key Results’” ¹. The talk plays a 1970s clip of Grove teaching it. Grove's own framing is telling: “The two key phrases of the management by objective systems are the objectives and the key results.” The objective is the direction; the key results “have to be measured, but at the end you can look and without any argument say, ‘Did I do that, or did I not do that?’ Yes. No. Simple” ¹.

So by Grove's own description, OKRs grew out of management by objectives, not out of nowhere. What Matters, summarizing Doerr's book, says Grove's original name for them was “iMBOs,” for Intel Management by Objectives ⁶; the management by objectives page covers that older system. The sources here do not fix the year Intel first used the name, and retellings that give one should be treated with care. What they do support is the part that matters for practice: the key result was meant to be a yes-or-no test, not a dial to monitor.

Doerr says he introduced OKRs to Google's cofounders in 1999 ¹. Google's internal guidance later split them in two. Committed OKRs are ones “that we agree will be achieved,” with an expected score of 1.0. Aspirational OKRs “express how we'd like the world to look,” and have an expected average score of 0.7, with high variance ². Both are about change. Neither is a gauge.

The mechanism: health and change need different machinery

The distinction is older than OKRs. Hewlett-Packard, whose standardized hoshin planning materials were influential, kept two documents: hoshin plans for a very few breakthrough objectives, and a separate business fundamentals table with more numerous KPIs for incremental improvement ³. Witcher's review of the literature also reports Nomi's analogy in Yoji Akao's edited volume: daily control is like paying attention to your health every day, with enough sleep and exercise; hoshin management is attending to critical events, such as an illness ³. Our page on hoshin kanri and the X-matrix shows how that split runs through a whole planning year.

The two need different machinery because a miss means different things. When a KPI leaves its normal range, the question is “what broke?” and the answer is usually operational. When a key result falls short, the question is “was the bet wrong, the plan wrong, or the resources wrong?” and the answer usually changes priorities. A What Matters explainer puts it plainly: KPIs most often monitor the “steady-state,” while OKRs describe outcomes ⁴.

• Put KPIs into an OKR set and you crowd out the few changes that matter with numbers that would have been hit anyway. The set scores well and the business stays where it was.
• Run a KPI like an OKR, with a stretch target and pressure to hit it, and you invite the cheapest way to move the number. Our page on KPI examples that track strategy covers that failure and the counter-measures that catch it.
• Run an OKR like a KPI, watched every week as a gauge with no plan behind it, and you get a wish with a chart.

Promote and demote: the life of a number

The practical rule comes from What Matters, the site that publishes Doerr's OKR material. KPIs can become OKRs “if it's a measurement that you want to change significantly,” and once the number is holding steady, “you can demote it to a KPI; you still want to keep an eye on it, but it doesn't need to be an OKR because it doesn't need to change” ⁵.

1. Watch it as a KPITrack the number against its normal range. No objective, no stretch target.
2. Promote it when it must moveWhen the number has slipped badly or the strategy needs it at a new level, write an objective around why it matters and key results that define the new level and the date.
3. Change somethingList what the team will do differently. If the list is empty, you have not promoted it; you have relabeled it.
4. Demote it when it holdsOnce the new level has held for a few periods, return it to the KPI set with a new normal range, and free the OKR slot for the next change.

Writing key results that are not KPIs or tasks

Most bad key results fail in one of two directions: they are KPIs (no change required) or tasks (activity, not result). Google's guidance gives the litmus tests ².

• The “who cares?” test. Could the OKR score 1.0 under reasonable circumstances without any direct user or economic benefit? If so, rewrite it around the benefit. Google's example: not “Launch X,” but “Double fleet-wide Y by launching X to 90+ percent of borg cells” ².
• The sufficiency test. Is it possible to score 1.0 on every key result and still miss the intent of the objective? If so, the key results are necessary but not sufficient; add or rework them until hitting them guarantees the objective ².
• The change test. What will the team do this quarter that it would not have done anyway? If the honest answer is nothing, it is a business-as-usual OKR. In Google's words, an OKR “that requires no changes to any group's activities is a business-as-usual OKR” ².

The case for one system, and what pay does to both

The strongest objection is that the distinction is jargon. A well-run balanced scorecard already has measures with targets, some held steady and some stretched. Why run two systems with two vocabularies, two software tools and two review meetings?

Fair, and one system is fine. The distinction does not need two tools; it needs two kinds of target. A target that defends a normal range and a target that requires change are different promises, and they need to be labeled differently in whatever system you use. Hewlett-Packard ran both inside one planning process, as two tables ³. What fails is a single list in which a defend-the-range number and a change-the-business number sit side by side with the same color rules, because the team will spend its effort where scoring is easiest.

Pay is where the distinction does most damage at scale. Doerr says that at Google, OKRs “are not used for bonuses or for promotions” ¹. The reason is mechanical. Tie a bonus to an aspirational OKR and it becomes a committed one: people set targets they know they can hit. Google's guidance warns that marking an aspirational OKR as committed “creates defensiveness in teams who cannot find a way to deliver the OKR,” and that a missed committed OKR calls for a postmortem that “is not intended to punish teams” ². If you link pay to anything, link it to KPIs with counter-measures and to committed OKRs, and keep aspirational OKRs out of the formula.

One more second-order effect. A What Matters advice column describes OKRs as a way to lead and KPIs as a way to manage ⁵. Senior teams that only review KPIs are managing well and leading nothing. Senior teams that only review OKRs are chasing changes while the base erodes. The quarterly review needs both, in separate parts of the agenda; the KPI dashboard template splits the screen the same way.

What to do Monday

Take your team's current OKRs. For each key result, write one sentence: what will we do differently this quarter to hit it? Move every key result with no answer to the KPI dashboard. Then look at the KPI dashboard and ask which number the strategy needs at a different level a year from now. That one is your next promotion.

A number does not know whether it is a KPI or a key result. You decide, by what you ask of it. Ask it to stay put, and watch it. Ask it to move, and change something, or admit you were only watching.

Questions

What is the difference between an OKR and a KPI?

A KPI is an ongoing measure of how part of the business is performing, watched against a normal range or standard. An OKR is a time-boxed commitment to change something: an objective that sets direction plus a few measurable key results. The quick test is whether anything must be done differently to hit the number. If not, it is a KPI.

Can a KPI be a key result?

Yes, when you need the number at a new level. Promote the KPI by writing an objective about why the change matters and a key result that sets the new level and the date, then change how the team works. When the number has held at its new level, demote it back to the KPI set and free the OKR for the next change.

Who invented OKRs?

John Doerr credits Andy Grove, under whom he was working in 1975. In a 1970s teaching clip shown in Doerr's 2018 TED talk, Grove describes objectives and key results as the two key phrases of the management by objectives system, which places OKRs as a descendant of management by objectives. The year Intel first used the name is not settled by those sources.

Should OKRs be tied to compensation?

Keep aspirational OKRs out of pay. Once a bonus depends on a stretch goal, people set goals they can be sure of hitting, and the aspirational OKR becomes a committed one. Google's guidance treats a missed committed OKR as a reason for a postmortem, not punishment. If pay must link to measures, use committed goals and KPIs paired with counter-measures.

Where to go next

• KPI Examples That Track Strategy, Not Activity
• KPI Dashboard Template
• Hoshin Kanri and the X-Matrix: Guide and Template
• OKR (Objectives and Key Results)
• Management by Objectives
• SMART Goals: Framework & How to Apply It