OKR

🗒️Objective#

No numbers, just the state you want to be standing in

Keep metrics out of the objective line. 'Cut checkout drop-off below 3%' is a key result, not an objective. The objective states in one sentence where the team should be standing when the quarter ends, something like 'nobody gets stuck the first time they pay us'. If reading it tells you what to stop doing, it is written well.

Strip the modifiers#

Delete the adjectives and read what is left

Delete every modifier from the draft and read only the nouns and verbs that survive. 'Best-in-class, rock-solid infrastructure' leaves you with 'infrastructure', which means there was no objective there. Put a concrete end-of-quarter scene in the gap. Show it to three teammates separately and see whether they picture the same scene; that is the whole check.

A cap on the count#

Three to five objectives a quarter, three to five key results each

Google's re:Work guide suggests three to five objectives per quarter, with three to five key results under each. The cap exists to force giving things up, not to keep the page tidy: the moment a sixth objective goes on the list, the decision about what not to do gets postponed. Long lists mean the team works the familiar items first and the rest sit unmet at quarter end.

📏Key Result#

Count what changed, not what you shipped

A key result has to be a sentence that comes out true or false at quarter end. 'Ship the onboarding redesign' is work done; 'day-7 retention for new signups at 34%' is a changed outcome. Score the work itself and you can ship, change nothing, and still take full marks. Write the baseline next to the number, because that is what tells you whether 34% is a rise or a fall.

Output metrics and outcome metrics#

Do not put them in the same box

'Two user interviews a week' is an output metric. The team controls it completely, so it is easy to hit and it gets filled in whether or not anything changed. Keep outcome metrics in the key result boxes and park output metrics in the execution plan. If you cannot produce an outcome metric at all, you have not yet decided what you are trying to change.

Guardrail metrics#

Hang the numbers that must not break, too

Next to the number you are pushing up, hang the numbers that must not break. In a quarter aimed at signup conversion, put the 30-day refund rate and inbound ticket volume on guardrail duty. Without them the team finds the cheapest way to move the metric, and the bill arrives next quarter. Guardrails are not scored; they are a pass condition.

🔗Alignment#

Not handed down from above, matched sideways

Slice the top objective and hand the pieces to lower teams and OKR turns into an assignment sheet. Alignment is each team bringing a draft it wrote itself and marking where that draft meshes with the objective above. Wherever a draft does not mesh, you have found something the organization has not agreed on yet. The output of an alignment meeting is that list, not an approval.

Vertical alignment#

A key result upstairs becomes an objective downstairs

One key result on the floor above drops into the objective slot on the floor below. Sentences get more concrete as you descend while the clock stays fixed to the quarter. Past three floors, the bottom can no longer see the sentence at the top. In a large organization it works better to stop at two floors and let each team resolve the rest internally.

Horizontal alignment#

Items that cannot finish without another team

If a key result cannot land without another team's schedule, the same line has to appear in that team's OKR too. A dependency written on one side only comes back as a priority fight halfway through the quarter. Before the quarter starts, pull the dependent items into one place and check that both sides wrote the same sentence. If they cannot match, drop that key result from this quarter.

🎚️Scoring (0.0 to 1.0)#

Not a report card, an input for next quarter

At quarter end each key result gets a value between 0.0 and 1.0. Google's re:Work guide treats an average of 0.6 to 0.7 on aspirational goals as the healthy range. If everything comes in at 1.0, the goals were set too low. What the scoring is actually good for is not the number but the one line recording why that number came out.

What 0.6 to 0.7 means#

Goals written to fall short on purpose

An aspirational goal is set knowing from the start that it will not be filled. So 0.7 is not a failure, it is the value the design predicts, and at 0.3 you check the plan before you blame the goal. The rule only holds while a team can report a low score safely. Once it cannot, next quarter's goals quietly come down.

Kept away from reviews and pay#

Scores do not touch compensation

Google's re:Work guide is explicit that OKR scores stay decoupled from performance reviews and compensation. The moment a score touches money, everyone writes only the goals they are sure to clear, and aspirational goals disappear. Run reviews off separate evidence, and read OKR scores only in the meeting where next quarter's goals get set.

⚖️Committed and aspirational#

Same format, different promise

A committed OKR is an operational promise to finish inside the quarter, so 1.0 is the bar and anything short gets a cause analysis. An aspirational OKR is allowed to come in unfinished, and in exchange it changes the board when it lands. Write both on one page without marking which is which and the scoring bar blurs, until nobody can tell whether 0.7 is praise or a warning.

What happens when they mix#

One letter of labeling changes the order

Put committed and aspirational items in one unlabeled list and the team fills the committed ones first, pushing the aspirational ones to the end of the quarter. Those pushed items end up untouched with a low score, and next quarter they drop off the list entirely. Marking each item with a single letter for its type is enough to change that order.

🗓️Running the quarter#

The three months after matter more than the writing day

OKR dies quietly in the three months after the writing meeting, not in the meeting. Write it at the start of the quarter, open it again at the end, and nothing records what the team was working from in between. Open it briefly every week for the numbers and the blockers, and the format stays a decision tool instead of a plan document.

Weekly check-in#

Fifteen minutes, numbers and blockers only

A weekly check-in covers three things: the current number on each key result, the change from last week, and where it is stuck. Let it turn into a status presentation and the meeting gets longer while no decision comes out. Three flat weeks in a row means nobody touched that key result this quarter, so decide right there whether to move resources or drop the item.

Closing the quarter#

What carries over matters more than the score

The part of the closing meeting worth time is not the scores, it is what happens to the items that fell short. Decide per item whether it carries over as is, gets rewritten, or gets dropped, and leave one line recording the call. An item carried three quarters in a row is not a goal, it is something the team has decided not to do.