Writing OKRs

Route by task

Write workflow

1. Structure the output using the template. Load references/03-templates.md and fill every field. Never omit Parent OKR, Type, or Why this matters.

1. Draft Objectives. Each must be qualitative (no numbers), directional, and pass the "Who Cares?" test: if we 1.0'd this, would a customer, investor, or employee notice? If not, reword or cut.

1. Draft Key Results using the 4-part formula:

   [metric name] from [baseline] to [target] by [date]

If a baseline is unknown, use an explicit ⚠️ placeholder — never silently omit it. Format: ⚠️ [replace with actual — e.g., 88%]. A KR with an unfilled baseline cannot be scored at mid-cycle.

1. Apply the two litmus tests to every KR:

• Output vs. outcome: Can the KR be checked off without a metric moving? If yes, it's a task — rewrite as an outcome.
• Necessary but not sufficient: If the team scored 1.0 on every KR, is it possible the Objective still wasn't achieved? If yes, the KR set is insufficient — add or rewrite KRs until their collective success guarantees the Objective.

1. Tag every OKR as committed or aspirational (see Committed vs. aspirational). An untagged OKR is a rubric hard fail.

1. Separate initiatives from Key Results. List initiatives (the projects and bets believed to move the KRs) in the Initiatives section — not in the KR list. If initiatives appear in KR slots, move them.

1. Final rubric pass. Run references/01-rubric.md on every Objective and KR. Include a compact rubric summary table in the output (dimension → score → pass/fail per Objective and per KR) so the quality check is transparent. Fix anything below threshold before finalizing.

Refine workflow

1. Read the draft OKRs in full.
2. Apply the two litmus tests (output vs. outcome; necessary but not sufficient) to every KR.
3. Check that every KR has all 4 formula parts: metric, baseline, target, date.
4. Check that every Objective passes the "Who Cares?" test and contains no numbers.
5. Check committed/aspirational tagging on every OKR.
6. Run references/01-rubric.md on each Objective and KR.
7. Produce per-OKR output:

• Compact rubric score table (dimension → score → pass/fail)
• Narrative critique and concrete rewrite suggestion for every failing dimension
• Clearly distinguish hard fails (any 0 on K1, K2, K8, or any Objective dimension) from point deductions

Evaluate workflow

For program-level audits, planning-week reviews, or end-of-cycle retrospectives.

1. Score each Objective on all 6 dimensions from references/01-rubric.md. Flag any 0 as a hard fail.
2. Score each KR on all 8 dimensions. Flag any 0 on K1, K2, or K8 as a hard fail.
3. Run the two OKR-as-a-whole checks: type tagging and initiative separation.
4. Check for anti-patterns using references/02-anti-patterns.md.
5. Produce:

• Rubric score table per OKR
• Narrative per failing dimension with specific rewrite suggestions
• Anti-pattern findings, grouped by severity
• Health verdict: healthy (0 critical, ≤2 high) / needs work (0 critical, 3–5 high) / at risk (any critical, or ≥6 high)
• Top 3 prioritized fixes

Committed vs. aspirational

Every OKR must be explicitly tagged. Mislabeling has predictable failure modes: treating an aspirational OKR as committed drives sandbagging; treating a committed OKR as aspirational signals it's optional.

Reasonable starting mix: two-thirds aspirational, one-third committed. Too many committed OKRs turns the program into a delivery contract.

Key constraints

• 3–5 Objectives, 2–4 KRs each. Start by drafting, then cut. Teams capping at 1–3 Objectives complete significantly more KRs than those juggling five or more.
• No numbers in Objectives. Numbers belong in KRs.
• KPIs are not OKRs. A KPI is a vital sign monitored continuously. A KR is a deliberate campaign to move a KPI from X to Y by date Z. A KPI becomes a KR when you commit to acting on it.
• Initiatives belong in the roadmap, not the KR list. Conflating them is the silent killer of most OKR rollouts.
• Binary KRs are task KRs in disguise. "Ship X" is binary: done or not done. Reframe as the outcome the ship is intended to cause.

Alignment

Every OKR should answer: what bigger thing does this support? Record it in the Parent OKR field.

• Use lineage (family tree) over strict cascading. Children can be proposed bottom-up; leadership doesn't pre-assign them.
• For cross-functional outcomes, use a shared Objective with joint ownership rather than splitting the outcome across team OKRs that pretend to be independent.
• Cap alignment depth at 2 levels (company → team) for most orgs. Three levels only when the org genuinely needs an intermediate layer.
• Healthy top-down/bottom-up mix: ~60% from leadership direction, ~40% from the team's own knowledge of the work.

Quick anti-pattern scan

Before finalizing any OKR set, check these four critical failures first:

Full 20-item checklist with severity ratings and standard fixes: references/02-anti-patterns.md