OKRs
OKRs
Objectives and Key Results is a goal-setting FRAMEWORK. The goal of OKRs is to create alignment and motivation around goals (cascade through the organization and helps in prioritization).
Objectives - What do we want to achieve? (must be ambitious, no ambiguity) KRs are definitions of success and failures for the objectives they are defined against. How do we identify if we have achieved the objective? KRs should be measurable and easy to grade (can set pre-scoring criteria)
Why do we need it? For situations like:
- I don’t know why am I doing this
- Everything is #1 priority
- How will we know when ae we done?
DOs
- Transparency is important - share with everyone in the team.
- Around 5 objectives with 3-5 KRs per objective at a time
- The sweet spot for the final score is 60-70% - if it’s more than 90 then maybe you’re not ambitious)
- KRs are outcome-oriented rather than output-oriented
Bad: Increase task CPU utilization by 3 percent
Good: Decrease quantity of cores required to serve peak queries by 3 percent with no change to quality/latency/ . . . and return resulting excess cores to the free pool
- For larger groups, make OKRs hierarchical—have high-level ones for the entire team, more detailed ones for subteams. Make sure that the “horizontal” OKRs (projects that need multiple teams to contribute) have supporting key results in each subteam.
DONTs
- Grading should not be linked to performance reviews/evaluations.
- If you wrote them down in five minutes, they probably aren’t good. Think.
- If your objective doesn’t fit on one line, it probably isn’t crisp enough
- Make sure the metrics are unambiguous. If you say “1 million users,” is that all-time users or seven-day actives?
Types -
Committed OKRs -
Agreed to be achieved and thus willing to adjust resources/re-arrange other priorities to ensure they get done (escalation essential)
The expected score is 1 - anything less means was not planned/executed properly (postmortem required)
Aspirational OKRs -
- How the best world would look like
- Not sure how to get there
- Expected score < 1 (high variance)
Cadence -
The frequency at which previous OKRs have to be graded and new ones to be created (usually quarterly)
- Annual OKR setting and review
- Quarterly review, grading, and changes if required
- Monthly updates (including blockers)
Scoring Guide -
Usually graded on a score of 0 to 1 (1 means fully achieved)
Committed OKRs should be scored either 0 or 1. (BInary - either done or not)
Aspirational OKRs can be scored b/w 0 to 1 (0.7 being the sweet spot, high variance)
For example: KR is “Improve x by 30% by end Q3 2022” so scoring can be
0.3 - 10% improvement, 0.6 - 20% improvement, 1 - 30% improvement
Each KR can have a monthly score b/w 0 to 1 - each month the score “should” increase (same for each quarter). Overall score can be the average of all KRs.
OKRs are big, not incremental — we don’t expect to hit all of them. (If we do, we’re not setting them aggressively enough.) We grade them with a color scale to measure how well we did:
- 0.0–0.3 is red
- 0.4–0.6 is yellow
- 0.7–1.0 is green
Source -