---
name: okr-leadership-review-author
description: "Load when a user pastes an OKR or goal-tracking export — objectives, key results, progress values, owners, due dates, and optionally trend or last-update timestamps — and asks for a leadership review narrative, quarterly business review write-up, board pre-read, or monthly check-in summary. Returns a structured single-screen page: executive summary leading with posture and the top decision, a leadership-decisions section (empty when none qualify), a blocked/at-risk table, per-objective narratives with named owner actions, and informational updates. Handles plain text, markdown, JSON object, and labelled-paragraph formats."
---

# The Ledger, Not the Lamp

Leadership reads an OKR review on its way into the room: the first screen closely, nothing after it. So the whole page fits on that screen, its verdict and first decision sit in the opening sentences, and every line below is a call the tracker could not have made.

The tracker hands over a lamp per key result: a colour, a percentage, a date, a line of trend. The owner lit it on the day of the update, with whatever hope the owner had that day. It says nothing about whether the definition moved, whether a nearer date governs, whether an input failed, or whether anybody but the owner checked the figure. The review reads the ledger instead. Every key result committed four terms at the start of the quarter, and every failure a review must catch is one of them breaking while the lamp stays lit.

## Pace against the clock that governs

Pace is the share of work done over the share of time used: progress ÷ (weeks elapsed ÷ 13). Use the export's own progress figure. At week 4, 30% done is 0.30 ÷ (4/13 = 0.308) = 0.97. At week 13 the time used is 1.0 and the ratio is the progress. Write every ratio with its inputs; one wrong division costs the page every right one.

About 0.9 or more is on-track. From 0.7 to 0.9 is at-risk: the owner can close it. Below 0.7 is off-track. At the close every result under 1.0 is not met, and the band says by how much. Early in a quarter low figures are normal: judge week 4 on pace, dates and dependencies, and say where the colour is more alarmed than the arithmetic.

The end of the quarter is only the default clock; it is not always the one in charge. A regulator's filing day, a queue that closes, a price that lapses, or a downstream key result that needs this one first sets an earlier date, and time runs to it: progress ÷ (weeks elapsed ÷ the week of that date). Work due at week 8 that is 20% done at week 4 reads 0.20 ÷ (4/8 = 0.50) = 0.40, not 0.20 ÷ 0.308 = 0.65. Put the governing ratio first. A comfortable ratio in front of a door about to shut is the most expensive mistake a review can make, so work backward from the date: what has to be true, and by which day.

## Settled, not paced

Filings, approvals, launches, certifications and inspections either happened or did not. A qualified outcome — a conditional rating, a provisional certificate, a pass with findings open — is not the result committed; the verdict is not met, the GREEN is overruled in those words, and the exposure is named. A threshold is met or missed on the last day: 97.1% against 98% is missed, however steeply it climbed; the trend is next quarter's starting point. A milestone that showed 0% right up to the week it landed gets full credit, plus one sentence noting that the updates never tracked the work.

Measure each key result against the words it was written in. A narrower delivery is a scope change, not a partial success, and accepting it is leadership's. Whatever was sized to the original inherits the miss. A key result needing another's output carries that one's risk whatever its own ratio. One outside body holding two key results is one blocker; one team wanted by two is one resourcing conflict. A key result planned to start after another finishes is sequencing, not failure: name the gate and its date. A figure untouched for five weeks, hand-entered, a trend word the owner typed rather than the system recorded, or "in final review" with nothing shared is a measurement gap, named beside the pace verdict.

Each owner, date, figure, currency and year is taken from the export. Where the export is silent, say so and name who can fill the gap.

Read `references/classification-guide.md` for the full decision checklist, the not-decisions list, sector clocks, evidence standards, and common misreadings before writing the decisions section.

## The page

Spend words on judgement, none on what the tracker already says: one sentence per at-risk key result, one or two per narrative, one line of updates. Order makes any cut safe: summary, every decision, at-risk entries, narratives, updates.

Read `references/page-template.md` for format rules by form (plain text, markdown, capitals, JSON) and the five-section outline, then `references/worked-examples.md` for five complete worked pages before writing.