What are the best practices for OKRs?
The rules that survive contact with reality
1/ Focus. Two to four objectives per team, two to four key results each. More than that and nothing is a priority. Less is more, and less is hard.
2/ Measure outcomes, not tasks. A key result is a result, not a to-do. "Launch the newsletter" is a task. "Grow active subscribers from 4,000 to 6,000" is an outcome. If it can go in your backlog, it is not a key result.
3/ Check in weekly. The rhythm is the whole game. See the cadence guide.
4/ Keep it off pay. Tie OKRs to bonuses and you get sandbagging. Run compensation on its own track.
5/ Aim high. Set OKRs so about 70 percent is a strong result. Hitting 100 percent every time means the goals were too safe.
6/ Align by negotiation. Direction from the top, goals from the teams. See the alignment guide.
7/ Grade and reflect. Score OKRs at cycle end and turn the score into a lesson, not a verdict.
What the canon actually says
These rules are not invented. OKRs came from Andy Grove at Intel, who took Peter Drucker's Management by Objectives and added measurable key results and a quarterly cadence. John Doerr carried the method to Google in 1999 and later wrote Measure What Matters. His mnemonic for what OKRs deliver is F.A.C.T.S: Focus, Alignment, Commitment, Tracking, and Stretching. Every rule above maps to one of those five.
Doerr pairs OKRs with a second practice he calls CFRs: Conversations, Feedback, and Recognition. The point is blunt. The black-and-white grading of OKRs means little without the human conversation around it. A weekly check-in is a CFR in disguise. This is also why OKRs and performance reviews should stay apart: the conversation should be about the work, not the rating.
The anti-patterns to name
Too many OKRs. Set-and-forget. Comp-linkage. Task lists dressed as key results. Rigid top-down cascade. Treating steady-state operations as OKRs. Each one is a known killer. See why OKRs fail.
The practice under the practices
None of this works if nobody checks whether it is happening. Best practices are only real if you can see adherence. That is the case for measuring program health: it turns a list of good intentions into a number you can act on.