The components, and what good looks like
| Component |
What good looks like |
The common corruption |
| Objective |
Qualitative, significant, time-bound: a destination worth the quarter |
A department name with a verb attached ("improve marketing effectiveness") |
| Key results |
Two to five measurable outcomes; a stranger could grade them from the data |
A task list: "launch the campaign" measures activity, not whether it worked |
| Cadence |
Set quarterly, reviewed frequently, scored and discussed at cycle end |
Set in January, rediscovered in December |
| Scoring |
Honest grading where ambitious targets scoring around 0.7 counts as strong |
Everything lands at 1.0, which means the targets were negotiated, not set |
| Scope |
The few things that must change this cycle, alongside business as usual |
Everything the team does, rewritten in OKR format |
Lineage in one paragraph
The approach descends from Andy Grove's management practice at Intel, where objectives were paired with the measurable "key results" that would prove them. John Doerr carried the method to Google in 1999, and its later spread across the technology sector turned it into a default. Two Grove-era principles are routinely lost in translation: key results measure outcomes rather than effort, and ambitious OKRs are deliberately decoupled from compensation so that stretch remains safe.
The decisions OKRs are actually good for
OKRs are a prioritisation instrument disguised as a goal format. They work best when an organisation has more credible initiatives than capacity and needs a mechanism that forces choices: which three outcomes matter this quarter, what will not be pursued, and how everyone will know whether it worked. Used this way, the quarterly OKR discussion becomes the executive team's recurring resource-allocation decision. That is why weak OKRs are usually a symptom of an unmade strategic choice rather than a formatting problem.
A short illustration
A scale-up's leadership drafts "Objective: strengthen our enterprise offering" with key results that are all launches (SSO shipped, security page published, sales deck refreshed). Every KR could hit while enterprise revenue stands still. The redrafted version keeps the objective and replaces the key results with outcomes: enterprise-tier logos signed, sales-cycle length for enterprise deals, expansion within existing enterprise accounts. Two of the original tasks survive, demoted to what they always were: initiatives that might move the key results, or might not. That demotion is the framework doing its job.
The corruptions, catalogued
- Key results written as tasks. Activity is graded, outcomes are not, and the team hits 100 per cent of its OKRs in a flat quarter.
- Sandbagging once pay is attached. Wire bonuses to OKR scores and every ambitious target disappears within two cycles, because rational people do not stretch against their own income.
- Priority sprawl: eight objectives with five key results each is not focus, it is the old operating plan with new headings. If everything is an OKR, the framework has selected nothing.
- Top-down cascading rebranded as alignment: leadership fixes the numbers and teams decompose them in a spreadsheet, removing the bottom-up commitment that makes people own the target.
- Grade inflation at review: misses are re-narrated as "directionally achieved", so the scoring loses its information content and next cycle's targets are set on fiction.
Where independent perspectives sharpen the cycle
The hardest OKR questions are not about format. They are the strategic choices hiding underneath: whether these are the right three objectives, whether the key results measure the business or flatter it, whether the stretch is honest. Internal debate on those questions is bounded by the same assumptions each quarter. Selected senior operators who have run comparable businesses can challenge the objective set itself before the cycle locks, which is worth more than any amount of KR wordsmithing after it does.
