OKR Reviews skill for Codex

A quarterly review is worth the time when scores are backed by data anyone can check, misses are explained by cause rather than excuse, each lesson changes something next quarter, and leadership leaves knowing what is asked of them. Reviews that grade effort instead of outcomes teach teams to sandbag.

Workflow

• Freeze the data: pick a cut-off date and pull every key result’s final value from its source (dashboard, query, tracker), with a link
• Owners self-score with evidence and a cause for each shortfall
• Calibrate scores across teams (same scale, same standards)
• Hold the retro per team: what worked, what did not, what changes
• Draft the QBR: scorecard, narrative, misses, risks, asks, next-quarter priorities
• Decide carry-overs explicitly (continue, change, drop) - feed into okr-writing
• Publish and archive with the source data links

Read the original OKR document first. Score against the key results as written at the start of the quarter; if a KR was formally changed mid-quarter, note the change and the reason beside the score.

Scoring

Use a 0.0-1.0 scale (or 0-100%) based on how much of the measurable target was achieved.

OKR type	Expected score	Interpretation
Committed (must deliver)	1.0	Anything below 1.0 is a miss needing explanation and a plan
Aspirational (stretch)	0.6-0.7 on average	Consistent 1.0 means targets were too safe; consistent 0.3 means overreach or poor execution

Linear scoring for numeric KRs: score = (actual - baseline) / (target - baseline), capped at 0 and 1. Example: activation baseline 32%, target 40%, actual 37% gives (37 - 32) / (40 - 32) = 0.625.

For milestone KRs (ship X by date), define partial credit before the quarter, or score binary. Never award points for activity (“worked on”, “started”).

Classifying misses

Every KR below its expectation gets one primary cause:

Cause	Signal	Typical response
Execution	Plan was sound; delivery slipped (capacity, dependencies, quality)	Fix planning, staffing or dependencies
Wrong bet	Delivered the work, metric did not move	Record the learning; change approach, not just effort
External	Market, regulation, partner, outage beyond control	Note it; consider leading indicators next time
Bad target	Baseline wrong, metric unmeasurable, target arbitrary	Fix measurement before setting the next target
Deprioritised	Leadership moved focus mid-quarter	Record the decision and who made it

“Wrong bet” is valuable information, not failure; aspirational OKRs exist to make bets.

Calibration

• Compare scores across teams for the same kind of KR. If one team scores milestone KRs generously, adjust or flag it.
• Challenge any 1.0 on an aspirational KR (was it a stretch?) and any score without a data link.
• Check for KRs that were achieved while the objective clearly was not (for example, shipped all features but customers did not use them). Comment on the objective, not just the KRs.

Retro format (45-60 minutes per team)

1. Scorecard walk-through (10 min): scores and causes, no debate on numbers that have sources.
2. What helped us hit targets? (10 min, silent writing then share)
3. What got in the way? (15 min, grouped into themes)
4. What will we change next quarter? (15 min): at most three changes, each with an owner.
5. Carry-over decisions (5 min): continue, change or drop each unfinished objective.

Facilitation techniques are in meeting-facilitation.

QBR document structure

# <Team/Company> - Q<n> <year> business review
**Headline:** <one sentence: the quarter in outcome terms>

## Scorecard
| Objective | Key result | Baseline | Target | Actual | Score | Type | Cause if missed | Source |
|---|---|---|---|---|---|---|---|---|

## Key metrics vs plan
| Metric | Plan | Actual | vs last quarter | Comment |
|---|---|---|---|---|
<revenue, customers, retention, delivery, hiring, burn - whatever the business runs on>

## Wins (3 max)
<Outcome, evidence, what made it work>

## Misses (all material ones)
<KR, cause, what we learned, what changes>

## Customer and market signals
<Patterns from customers, competitors, sales; with examples>

## Risks for next quarter
| Risk | Likelihood | Impact | Mitigation | Owner |

## Next quarter priorities
<3-5 objectives, draft; link to OKR doc>

## Asks
<Decisions, budget, headcount, cross-team help - each with who and by when>

For a board or customer-facing QBR, keep the scorecard and asks; move team-level detail to an appendix. Customer QBRs (vendor to client) belong to customer-success-playbooks.

Writing the narrative

• Lead each section with the conclusion, then the evidence.
• Report misses with the same prominence as wins. Leaders trust reviews that show bad news early.
• Use numbers with comparison: vs target, vs last quarter, vs same quarter last year.
• Lessons must name a change (“we will set targets only on metrics with 8 weeks of baseline data”), not a sentiment (“communication could improve”).

Gotchas

• Rewriting KRs late in the quarter to match what was achieved; it destroys the signal.
• Scoring from memory or optimism the night before; freeze data and link sources.
• Too many KRs (more than 3-4 per objective) makes the review a list-reading exercise.
• Average scores across committed and aspirational OKRs hide missed commitments.
• Retros with no follow-up: check last quarter’s “what will change” items first.
• Treating OKR scores as individual performance ratings encourages sandbagging.

Output

Scored OKR table with sources, retro notes with up to three owned changes, carry-over decisions, and the QBR document (or deck outline for pptx-presentations) with asks.