OKR Writing skill for Codex

An OKR set is done well when the objective says why the quarter matters in words people remember, each key result is an outcome with a baseline, target and data source, nobody could hit every key result by completing a task list, and the set is small enough that the team can name it from memory.

Anatomy

Part	Rules	Bad	Good
Objective	Qualitative, directional, inspiring, time-bound by the cycle	“Q1 product goals”	“Make the first week feel effortless for new teams”
Key result	Outcome, number, baseline, target, source	“Launch new onboarding”	“Raise 7-day activation from 22% to 32%”
Initiative	Work you believe will move a KR; lives outside the OKR	(listed as a KR)	“Redesign invite flow; Slack invite; templates”

Limits: 1-3 objectives per team per quarter, 2-4 key results per objective. More than that dilutes focus.

Committed vs aspirational

Label every OKR:

• Committed: the team expects to hit it fully (score 1.0). Missing it needs an explanation. Use for things the business depends on (reliability, compliance, launch dates with external commitments).
• Aspirational: set so that about 0.6-0.7 is a good result. Missing the full target is normal.

Mixing the two without labels means leadership reads every 0.7 as a failure, or every 1.0 as sandbagging.

Turning outputs into outcomes

For each task-like KR ask “what would be different for customers or the business if we did this?” and measure that.

Output KR (rewrite)	Outcome KR
Ship SSO	Close 5 enterprise deals that listed SSO as a blocker (from 0)
Write 10 help articles	Cut “how do I” tickets from 340/month to 200/month
Migrate to new billing system	Reduce failed payment churn from 2.1% to 1.2% of MRR monthly
Run 8 A/B tests	Lift trial-to-paid conversion from 9% to 11%

Some KRs are legitimately binary (a regulatory deadline). Allow at most one per objective, label it committed, and add the outcome it protects in a note.

Key result checklist

• Starts with a verb of change: raise, cut, reach, reduce, grow
• Has a baseline (or “baseline unknown - measured by week 2”)
• Has a target and a cycle end date
• Names its data source or dashboard
• The team can influence it substantially within the quarter
• Cannot be hit by gaming alone; a guardrail is noted if it could
• Not a restatement of another KR in the set

Alignment

• Each team objective cites the company or department objective it supports.
• Do not copy a parent KR down as a child objective word for word; translate it into what this team can move (“Company: grow self-serve revenue 30%” becomes “Onboarding team: raise 7-day activation 22% to 32%”).
• Shared KRs across teams are fine; name one accountable owner team.
• Some OKRs may be bottom-up (platform health, developer experience). That is expected.

Workflow

- [ ] 1. Read company/department OKRs, strategy, last quarter's scores
- [ ] 2. Gather metric baselines (see $product-metrics); mark unknowns
- [ ] 3. Draft 1-3 objectives in plain language
- [ ] 4. Draft 2-4 KRs each; run the KR checklist
- [ ] 5. Rewrite output KRs as outcomes; move outputs to initiatives
- [ ] 6. Label committed vs aspirational
- [ ] 7. Check alignment links and capacity against the roadmap
- [ ] 8. Present for review; note open questions on baselines

Scoring at quarter end

Score each KR 0.0-1.0 by progress from baseline to target: score = (actual - baseline) / (target - baseline), clamped to 0..1.

Score	Committed meaning	Aspirational meaning
1.0	Expected	Target was likely too low
0.6-0.9	Miss; explain	Good result
0.3-0.5	Significant miss; root cause	Partial progress; learn why
< 0.3	Escalate and re-plan	Wrong bet or wrong metric

Objective score = average of its KRs (unless the team agreed on weights). Every KR score needs one or two sentences: what happened, what was learned, carry over or drop. Scores inform planning; they are not performance ratings for individuals. For a full quarterly business review or retro format, use $okr-reviews.

Output template

## <Team> OKRs - Q1 2027
Supports: <company objective>

### O1: <objective> (aspirational)
| KR | Baseline | Target | Source | Owner |
|---|---|---|---|---|
| KR1 Raise 7-day activation | 22% | 32% | Looker: Activation v2 | Onboarding |
| KR2 Cut median time to first shared doc | 2.4 days | 1 day | events: doc_shared | Onboarding |
Guardrail: support tickets per 100 new teams stays <= 6
Initiatives (not KRs): invite redesign, Slack invite, starter templates

### O2: ...
Open questions: baseline for KR3 unknown; data team to confirm by Jan 10

Gotchas

• A KR that only measures activity (“hold 20 customer calls”) rewards busyness.
• Too many committed OKRs remove room for ambition; too many aspirational ones make planning unreliable. One or two committed per team is typical.
• Changing targets mid-quarter hides learning; if a KR is clearly wrong, retire it and write down why.
• Business-as-usual work (on-call, support) does not need an OKR unless it is changing; reserve capacity for it in the roadmap instead.
• Never invent baselines. An OKR with a fabricated baseline produces a fabricated score.

Verification

Run the KR checklist on every key result; confirm no KR could be fully achieved by shipping a listed initiative without any metric moving; confirm each objective has an alignment link and a committed/aspirational label; and when scoring, recompute each score from baseline, target and actual.