The Test Bench240 guides, 20 benches
Conversion.sale
The Test Bench - a field guide to conversion for small online businesses
Home / Bench 20 - Audits, checklists and mistakes / Guide

Audit Output: Turning Findings Into a Fix List People Follow

An audit fails at the handover: findings die in a forty-page document nobody reopens. The output that works is a ranked list - each item with its cause, evidence, cost, owner and read rule - plus the deferred list with its reasons. This guide sets the format.

Published 24 October 2026Evidence checked on 24 October 2026Small online businesses, global English edition

The short answer

Publish the audit as a list, not a document. Each item carries: the finding in one sentence (stage and cause), the evidence in one line (what was seen, counted or quoted), the repair proposed, its cost in hours, its rank from the prioritisation score, and its read rule - how and when the repair will be judged. Followed by the deferred list: findings consciously not scheduled, each with its reason. The list fits on two pages, is reviewed at the monthly sitting, and is the only version of the audit that exists operationally; the funnel file stays as the evidence archive behind it.

The item format, field by field

Finding sentence first, because items get quoted in isolation and must survive it: 'Mobile checkout starters abandon at the delivery step; the date range cannot be matched to their deadline' travels; 'checkout issues' does not. Evidence line second, with counts ('9 of 20 recordings; 6 of 14 delivery tickets in March') so a future reader can weigh it without reopening the file. Repair third, concrete enough to estimate. Read rule last and non-negotiable: the segment, the stage rate, the window, the success threshold. An item missing its read rule is not scheduled; it returns to the findings file until it can be judged.

Ranking visibly, so the order can be argued with

Show the score components beside the rank - leak size, evidence strength, cost - rather than a bare order. Visible scoring lets the business argue productively ('that leak estimate is double-counting the seasonal dip') and lets next quarter's re-ranking reuse the work. Where a lower-scored item is promoted for strategic reasons (a launch, a partner commitment), the promotion is written on the item. Hidden overrides are how fix lists lose credibility; an honest override, recorded, costs nothing and keeps the system believed.

The deferred list is half the output

Every audit finds more than a quarter can fix. The deferred list records the rest with reasons - cost, evidence pending, seasonality, platform limits - and a revisit trigger ('when mobile traffic passes X', 'after the theme upgrade'). Without it, deferred findings resurface as fresh urgencies in every meeting and podcast, and the audit's ranking work is redone monthly from memory. With it, the answer to any 'why are we not doing Z' is a line in a document, and the quarter's focus survives contact with the business's enthusiasm.

Owners and dates, even for a team of one

Each item gets an owner and a target window, in a one-person business as much as in a team: the owner is who verifies the read, the window is when it ships. Items age on the list visibly; an item deferred twice gets re-examined rather than re-copied (its evidence may have rotted, its cost may have changed, its stage may have been repaired by another item's work). The monthly sitting's whole agenda is this list: shipped items' reads, open items' progress, deferred items' triggers checked. Fifteen minutes, one document, no slides.

Closing the loop into the log

When an item ships, it becomes a change log entry with the audit item referenced; when its read completes, the verdict returns to the list and the item is struck through, not deleted. A year of struck-through lists is the audit practice's track record - the evidence, when someone proposes skipping the process, that the process pays. Keep the lists with the funnel files; together they are the business's conversion memory, and memory, as this whole bench argues, is the asset that compounds.

An item, written properly

Illustration of the format doing its work. Finding: 'Mobile product viewers who need the delivery date cannot find it; the date appears only after checkout start.' Evidence: 'Step rate: product view to add, mobile 9 percent against desktop 16; 7 of 18 mobile recordings show the delivery hunt; 11 tickets in the period ask the delivery question pre-purchase.' Repair: 'State the dated delivery estimate on the product page beside the price, from the existing delivery table.' Cost: '2 hours, theme edit.' Read rule: 'Mobile add rate over four whole weeks post-change against the baseline segment; success is a rise of 2 points or more held across the window.' One paragraph, and a stranger could run the repair and judge it. That is the standard every item on the list is held to.

Sources and evidence on the record

Evidence basis: Method guide. Any figure quoted is attributed to a named published source listed below; worked examples are labelled illustrations with round numbers, not measurements. Applies when: At the end of every audit, of any size

Next on the bench

Bench 20 - Audits, checklists and mistakes CRO Audits, Checklists and Mistakes: The Working Guide An audit is a disciplined inspection: baseline first, evidence per finding, findings converted into a prioritised fix list, and the list verified after repair. This bench holds the audit methods, the checklists that run them, and the mistakes catalogue that keeps a small business from paying tuition twice. Bench 20 - Audits, checklists and mistakes The Full-Day Audit: Evidence, Recordings and the Funnel File The deep audit spends a day building the funnel file: segmented rates, a recording sample per weak stage, visitor voice from surveys and tickets, and findings written to the standard the fix list can act on. This is the method, in order, with the time boxes. Bench 01 - Foundations and measurement Choosing What to Fix First: A Prioritisation Method for Small Stores You cannot fix everything, and the order matters. A simple scoring method sized for one person's time: size of the leak, certainty of the evidence, and cost of the fix, in that order of honesty.