OKRs Aren't Dead. They're Misused.

The framework isn't broken. The way you've been running it is.

You ran a quarterly planning session. Six hours, four people, one whiteboard. You walked out with three Objectives, nine Key Results, and a Notion page that everyone said they'd reference weekly.

Twelve weeks later you do the retrospective. The team hit 78% of the KRs. By every dashboard signal, the quarter was a success. Revenue is also flat. The strategic position you wanted to build hasn't moved. The new product surface you wrote into the Objective has not, in any meaningful way, shipped.

You sit with the data and the same thought lands as it lands every quarter.

OKRs don't work.

Wrong diagnosis. OKRs are not the problem. The way most teams run OKRs is the problem, and the failure mode has a name, and the name has been documented since 1975, and almost nobody who runs an OKR cycle has read the original literature on what they're actually supposed to do.

The framework is fine. You've been running it as if it's a target system. It was never a target system. It was a direction system. The two systems break in completely different ways, and the way yours has been breaking is exactly the way every misused direction system has broken since metrics were invented.

What John Doerr actually said

In 1999 a Kleiner Perkins partner named John Doerr walked into a meeting at Google with Larry Page and Sergey Brin and gave them an 80-slide deck about a goal-setting framework he had learned at Intel under Andy Grove. The framework was OKRs. Page and Brin adopted it that quarter. Google used it through every stage of growth. By 2017 Doerr had finally written the framework down for everyone else, in a book called Measure What Matters.

The book is the most-cited primary source on OKRs. It is also a book most operators have not actually read.

The chapter that gets quoted is the one with all the case studies. The chapter that gets ignored is chapter five, where Doerr is explicit about what OKRs are and aren't.

OKRs, Doerr writes, are not performance reviews. They are not bonus drivers. They are not used to compensate people. They are not used to evaluate people. The moment you tie them to compensation, the system breaks, because people start sandbagging the KRs to make sure they hit them.

The point of an OKR, Doerr keeps saying, is to align the team around a direction and to make the gap between the current state and the desired state visible. The standard at Google was that hitting 70% of your KRs was excellent. If you were hitting 100%, you were setting them too low. The 30% miss was the feature, not the bug. The miss was the data.

Now look at how you've been running yours.

You set the KRs. You measured the team against them. The team that hit 100% looked good in the all-hands. The team that hit 65% looked bad. People on both teams adjusted their behavior next quarter. The good-looking team set safer KRs. The bad-looking team learned to phrase their KRs more conservatively. Within two cycles, every KR in the company was something the team was 90% sure they could hit.

The OKR system was technically running. The OKR system had also been quietly converted into a slow-motion sandbagging exercise, because the moment you used the KRs to evaluate the people, the people started managing the KRs instead of the work.

That's not OKR failure. That's the predictable, well-documented failure mode of any metric used as a target.

What Charles Goodhart proved in 1975

In 1975 a British economist named Charles Goodhart, then advising the Bank of England, published a paper on UK monetary policy. The paper is now mostly forgotten. The single sentence inside it, paraphrased and sharpened by other economists in the years that followed, is one of the most-cited claims in social science.

"When a measure becomes a target, it ceases to be a good measure."

The paraphrase is now called Goodhart's Law. It applies to every metric system in every domain that has ever been studied. Test scores. Hospital wait times. Police arrest counts. Customer support response time. Code coverage. NPS. The pattern is universal.

The moment you tie incentives to a metric, people optimize the metric instead of the underlying outcome the metric was supposed to track. The metric goes up. The outcome stays flat or gets worse, because the energy that used to go into the outcome now goes into managing the metric.

Goodhart's Law is exactly what's happening to your OKRs.

The KRs are the metrics. They were originally instrumental, designed to track progress toward the Objective. The moment the team realized hitting the KRs would make them look good and missing them would make them look bad, the KRs stopped tracking progress and started getting managed. The Objective stopped moving, because the energy moved to the KR layer.

You can't fix this by writing better KRs. The fix is structural. The fix is the part of Doerr's book that everyone skips. Untie OKRs from evaluation. Aim for 70% as the goal. Treat the miss as data. Make the system about direction, not about score.

Most teams won't do this because it requires giving up the comforting illusion that the metric is the truth.

The metric is never the truth. The metric is, at best, a fingerprint of the truth. The moment you treat it as the truth, it stops being a fingerprint.

What W. Edwards Deming was screaming about in 1982

W. Edwards Deming was an American statistician who, after being ignored by American manufacturing, went to Japan in the 1950s and helped engineer the Toyota Production System. By the 1980s, when Japanese manufacturing had started to eat the American auto industry alive, American executives finally started reading him.

In 1982 Deming published Out of the Crisis. The book contains his famous 14 Points for Management. Point 11, repeated in every speech he gave for the last fifteen years of his life, is the one most operators have never heard.

"Eliminate management by objective. Eliminate management by numbers, numerical goals. Substitute leadership."

Deming had spent forty years watching organizations destroy themselves with poorly-implemented metric systems. The pattern was always the same. Management set numerical targets. The targets were either set too low (in which case they didn't drive change) or too high (in which case people gamed them). Either way, the targets pulled focus from the actual system that produced the outcome.

Deming's argument was that the variation you see in any system's output is mostly determined by the system itself, not by the people inside it. Setting harder targets for the people doesn't change the system, it just makes the people miserable until they game the targets. The fix is to redesign the system. The targets are downstream of the system.

Most senior operators read this and feel attacked. Their entire identity has been built on hitting numbers. They've been promoted, paid, and recognized for hitting numbers. The idea that the numbers are mostly noise from a system they didn't build is uncomfortable.

It's also accurate. Forty years of replication has not dented Deming's claim.

The implication for your OKR cycle isn't that you should stop running them. It's that you should run them with a clear-eyed understanding of what they can and can't do. They can give a team a shared direction. They can make the gap between current state and desired state visible. They can structure a quarterly conversation about what to work on.

They cannot motivate people. They cannot replace leadership. They cannot, on their own, change the underlying system that produces the outputs you're trying to track. If you're using them for any of those things, you're using them for things they were never designed to do, and the quarterly disappointment you keep feeling is the receipt for the misuse.

Three rules to run OKRs the way they were designed

You don't have to abandon the framework. You have to run it the way the people who designed it ran it.

1. Untie OKRs from compensation and evaluation. Doerr's rule, in writing, in chapter five of his book. The moment KRs become bonus drivers, sandbagging starts. Run the OKR review separately from the performance review. Tell the team explicitly that hitting 70% of KRs is the goal. Treat misses as data, not as failures.

2. Set Objectives that point in a direction, not at a number. "Become the default tool for X" is an Objective. "Increase MRR to $250K" is a target. The Objective is what the framework was built for. The target belongs on a different document. Targets are useful for execution. They are bad as Objectives, because they don't tell the team what to do when conditions change.

3. Run a mid-quarter check that's about the system, not the score. Six weeks in, run a 30-minute review. Don't look at the KR percentages. Look at what's working in the system that produced this quarter's progress. What's broken. What needs to change in the system, not in the team's effort level. This is the Deming move, applied at OKR cadence. The system change you make in week six is worth more than the effort change you ask for in week ten.

If you want a second voice in the room while you reset your operating cadence, that's exactly what a Build Audit does.

The reframe

Your OKRs aren't dead.

You've been running them as a target system, on a framework that was designed as a direction system, and you've been getting target-system failure modes. Goodhart's Law. Sandbagging. Metric-management theater. None of those are OKR-specific. All of them are what happens to any metric used the way you've been using yours.

The framework still works. You just have to actually run it the way Doerr wrote it, the way Deming begged you to, the way Goodhart's Law has been telling you for fifty years.

Untie the comp. Set Objectives that point. Review the system, not the score.

The OKRs do their job when the people running them stop asking them to do something else.