Developed by Andrew S. Grove at Intel (High Output Management, 1983) and carried to Google and beyond by John Doerr (Measure What Matters, 2018).
Sets a few qualitative objectives, each with three to five measurable key results, shares them openly across the organization, and grades them every quarter.
Nine draft key results for one onboarding objective, sorted by whether they can be graded
The objective: new customers are live and confident within their first month. Each tag is a draft key result. Select one, then the box for whether it states a result or a piece of work, and whether a number makes it gradable: only the top-right box holds key results. The first is placed as an example. Illustrative.
- Not placedPlacedKeep as a key resultPut a number on itMove it to the planA task at most
- Not placedPlacedKeep as a key resultPut a number on itMove it to the planA task at most
- Not placedPlacedKeep as a key resultPut a number on itMove it to the planA task at most
- Not placedPlacedKeep as a key resultPut a number on itMove it to the planA task at most
- Not placedPlacedKeep as a key resultPut a number on itMove it to the planA task at most
- Not placedPlacedKeep as a key resultPut a number on itMove it to the planA task at most
- Not placedPlacedKeep as a key resultPut a number on itMove it to the planA task at most
- Not placedPlacedKeep as a key resultPut a number on itMove it to the planA task at most
- Not placedPlacedKeep as a key resultPut a number on itMove it to the planA task at most
What it describes
Can be graded from 0 to 1 at quarter end
Your sort matched 0 of 9.Your sort matched 1 of 9.Your sort matched 2 of 9.Your sort matched 3 of 9.Your sort matched 4 of 9.Your sort matched 5 of 9.Your sort matched 6 of 9.Your sort matched 7 of 9.Your sort matched 8 of 9.Your sort matched 9 of 9.
- Keep as a key result
- A result with a number: grade it from zero to one at quarter end.
- Put a number on it
- The right result, but nobody could grade it. Add a number and a date.
- Move it to the plan
- Counted work, worth doing: it belongs in the plan that serves the key results.
- A task at most
- Neither a result nor measured: put it on a task list, or drop it.
The sort
- Median go-live 21 days: Keep as a key result. A result customers feel, with a number: at quarter end the median is either 21 days or it is not.
- Run 12 webinars: Move it to the plan. Easy to count and fully in the team's hands, which is why it is work. Twelve webinars can run while customers still stall.
- 80% use 3 features: Keep as a key result. A result that shows customers are actually using the product, with a number and a deadline.
- Customers feel confident: Put a number on it. The heart of the objective, with no way to grade it. It became onboarding satisfaction above 4.5 out of 5.
- Launch the welcome kit: A task at most. A one-time piece of work with no measure of what it changes.
- Early tickets down 30%: Keep as a key result. Fewer tickets in the first month is a result: customers stuck less often.
- Hire two leads: Move it to the plan. A number, but a number of things the team does. The hires serve the key results.
- Improve the docs: A task at most. Work, and no measure of done. If the docs matter, the ticket key result will show it.
- Satisfaction above 4.5: Keep as a key result. The rewritten version of feeling confident: a result, a number, and gradable.
An example
The onboarding team at a business software company drafted its quarterly objective as new customers are live and confident within their first month, and listed seven key results under it. The head of customer success read them in the planning review and asked one question of each: at the end of the quarter, how would we grade it from zero to one?
Four could not be graded as results. Run twelve onboarding webinars and hire two onboarding leads were work the team would do, easy to count, and they moved to the plan. Launch the new welcome kit was a task. Customers feel confident was the right result with no number on it, so it became onboarding satisfaction above 4.5 out of 5. The final set had four key results: median time to go-live of 21 days, eighty percent of accounts using three core features by day thirty, month-one support tickets down thirty percent, and the satisfaction score.
At quarter end the team graded them 0.9, 0.6, 0.7, and 1.0. The weakest sent next quarter's effort to adoption.
Why it works
An objective says where the team is going in words people remember. Key results say how everyone will know it got there, in numbers nobody can argue about later. The pairing keeps ambition and honesty in the same sentence: the objective can be bold because the key results will show exactly how far it got.
The discipline that matters most is keeping activities out of the key results. A list of things the team will do says nothing about whether customers are better off; a team can run every webinar and still have customers who never go live. Writing results makes the team commit to an outcome and leaves it free to change the work when the work is not moving the number. Grading each result from zero to one at the end of the quarter, apart from anyone's performance review, makes the miss information instead of a threat.
The thinking behind it
Andy Grove Andy Grove, Intel's third employee and later its chief executive, set out the objectives and key results system in High Output Management (1983), a manager's handbook drawn from his practice at Intel. John Doerr, who learned the system in a course Grove taught at Intel in 1975, carried it to Google and called Grove the father of OKRs in Measure What Matters (2018). The system pairs each objective with the measurable results that will show, at the end of the period, whether it was reached and how far the team got.
How to run it
- Write the objective in words. Name where the team wants to be by quarter end, in a sentence people will remember.
- Write three to five key results. Each one a measurable result with a number and a date, which together would show the objective was met.
- Test each with the grading question. Ask how you would score it from zero to one at quarter end; if you cannot, rewrite it.
- Move activities to the plan. Webinars, hires, and launches belong in the work that serves the key results.
- Align with the teams around you. Read the OKRs above and beside yours and remove conflicts before the quarter starts.
- Publish them. Make every team's OKRs visible to the whole organization.
- Check weekly and grade at the end. Track progress every week, grade each result at quarter end, and keep the grades out of performance reviews.
Where it goes wrong
The key results are a to-do list. Run webinars, launch the kit, and hire two people are all completed, every grade is 1.0, and customers are no better off. Spot it when nothing in the list could fail while the work succeeds; rewrite each as the result the work is meant to produce.
Grades feed bonuses. Teams then set targets they know they can hit, and the ambition the method exists for disappears. Keep grading separate from pay and treat 0.7 on a stretch result as good work.
The set grows to a dozen key results per objective. Nobody can hold that many in view, and effort spreads thin. Keep three to five, and cut the ones that would not change a decision.
When to use something else
- KPI Design. When the task is a standing set of measures for running the work, with guardrails and a decision tied to each threshold.
- Hoshin Kanri. When a few breakthrough goals must cascade through every level of a large organization over a year or more.
- Goodhart-Aware Metric Selection. When the worry is that a key result will be gamed once teams are judged on it.
Where it sits in SOLVE eX
Go to the source
- High Output Management, Random House, 1983
- Measure What Matters, Portfolio, 2018