The Quarterly Review: Baseline, Log and the Next Quarter's List

The quarterly review does four jobs in order. Refresh the baseline (the new quarter's before-picture, by segment). Read the whole change log, hunting patterns across entries rather than judging them one by one. Collect the audit list's verdicts: which repairs verified, which were noise, which leaks closed, which were merely quieter. And draft next quarter's list from the three documents together - new evidence, patterns, open leaks - ranked by the prioritisation method. Skipping the review does not pause the programme; it reverts it to opinion with better stationery, because the compounding lives in the pattern-reading, not in the individual fixes.

Close the quarter's baseline and open the new one: same definitions, same segments, the distortion calendar updated with the quarter's actual events (including the ones nobody scheduled). Where a kept repair permanently changed a stage, the new baseline absorbs it - that is what progress looks like in this system, and the review should say so explicitly, stage by stage, so the business can see its own compounding written down. Baselines are also where tracking changes are confessed: a new analytics setup, a consent change, a platform migration - each noted, each making future comparisons honest about their seams.

Individual log entries answer 'did it work'; the quarterly read answers 'what works here'. Patterns to hunt: change types that repeatedly pay (copy over layout, mobile over desktop, information over persuasion), stages that repeatedly resist repair (a diagnosis problem, not an effort problem), hypotheses about your buyers confirmed or refuted twice, and the programme's own biases - the benches never visited, the segments never split. Write the patterns as findings with the entries cited; they steer next quarter more reliably than any new external idea, because they are the only research ever conducted on your exact funnel.

Every repair promised a read; the review collects. Verified wins are struck and celebrated in one line each. Honourable losses - the repair that did nothing, the leak that was misdiagnosed - get their analysis now, while the evidence is findable: wrong cause, weak repair, or a read corrupted by an event. Unread repairs are the review's scandal list; each leaves with a date or a formal abandonment, because a programme that lets reads slide is training itself to ship changes and never learn, which is the most expensive habit in this entire field.

Draft the quarter's priorities from the documents, in the sitting: the open leaks re-scored with the quarter's new rates, the patterns' implications (more of what pays, a diagnosis block for what resists), the calendar's demands (the pre-peak audit, the seasonal preparations), and at most one exploratory item - a new bench, a new segment - chosen deliberately. The list leaves the room ranked, owned and dated, with its deferred items recorded. If the sitting cannot produce the list from the documents, the documents are the finding: the quarter's measurement failed somewhere upstream, and repairing that is item one.

The output is a single page filed with the log: baseline movement by stage, patterns found, verdicts collected, next list as ranked. One page, because a review that produces a deck produces a performance; a page produces a record. Four pages a year is the programme's autobiography - the document a future owner, partner or buyer reads to learn what this business knows about its customers. Write it like it will be read by a stranger who respects evidence. It will be.

Run the sitting in the order of the four jobs with rough clocks: baseline, twenty minutes; log patterns, twenty-five; verdicts, twenty-five; next list, twenty. The clock matters because the comfortable jobs (admiring the baseline) expand into the uncomfortable ones (collecting the verdicts) if allowed. Keep the same chair and the same documents each quarter; ritual is what lets a ninety-minute sitting do a day's work. And hold the sitting even in the quarter when trading was dreadful - especially then, because the log read whole is the fastest way to separate what the market did from what the business did, which is the quarter's only useful question.