What a Design Audit Report Should Contain: Full Example

We cannot show you a real Clarity Sprint report. So we invented a company, gave it a leaky trial, and wrote the report anyway.

A Clarity Sprint report puts one metric against a named category benchmark, ranks every fix by effort with a conservative ARR figure next to it, and ends with the scope of the next phase. It is a written document mapped against your own analytics, and you keep it whether or not you continue.

Every real report we write is under NDA. The one below follows the same structure, on an invented product with invented numbers. Only the benchmark is real. The offer itself is on our Clarity Sprint page.

Everything about “Tallyroom” is fictional. It is not a client, a composite of clients, or a real company.

The invented product

Tallyroom is a B2B expense-approval tool for companies with 50 to 500 staff. It runs a 14-day trial with no card.

Illustrative inputs:

Section 1 of the report: the metric against its benchmark

The report opens with one number and the row it belongs to.

Benchmark: a free trial with no card converts at 4% to 6% when it is good and 10% to 15% when it is great (ChartMogul Conversion Report, February 2026, 200 B2B companies).

Tallyroom sits below the good band. The target in the report is the top of that band, 6%, not the great band. A first target should be reachable inside one round.

A fix rarely closes a whole gap. The report assumes 30% to 50%:

That range is the budget ceiling for everything that follows. Nothing in the report is allowed to cost more than the gap it closes.

Section 2: where the trial loses people

Step	Trials reaching it	Share of signups
Signed up	1,200	100%
Connected an accounting system	540	45%
Created a first expense policy	300	25%
Routed a first approval	228	19%
Paid	38	3.2%

The largest single loss is the second step. 55% of trials never connect an accounting system, and the product shows nothing useful until they do.

Section 3: findings, ranked

Every finding carries its evidence, an effort estimate and a conservative annual figure. The figure is calculated from a conservative lift in trial-to-paid points.

#	Finding	Evidence	Effort	Conservative lift	Annual impact
F1	The accounting integration comes before any value. Move it after a first policy built on sample data	55% drop at step 2; recordings show users leaving to find API keys	3 weeks	+0.6 points	$414,720
F2	Approval emails do not show the amount or the requester, so approvers open the app to find out, or ignore it	24% of routed approvals wait more than 3 days; support tickets cluster on “what am I approving”	1 week	+0.3 points	$207,360
F3	On a 13-inch laptop, the recommended plan sits below the fold on the pricing page	Scroll maps; most trial users who visit pricing never scroll past the first plan	1 week	+0.2 points	$138,240
F4	Empty dashboard after signup	Fixed by F1, not priced twice	n/a	n/a	n/a
F5	Settings use three different button styles	Real, cosmetic, no measurable effect on the funnel	n/a	n/a	Not recommended now

The math for F1, as it appears in the report:

F1 to F3 together: +1.1 points, or $760,320 a year. That sits inside the conservative range from section 1, which is the check every report has to pass.

F5 is in the report on purpose. A finding with no effect on the number is still a finding. It goes at the bottom, marked “not now”, so nobody spends a sprint on it.

Section 4: what AI did in this sprint, and what it didn’t

This section is also illustrative, but it describes how we work.

AI summarised the session recordings, clustered several hundred support tickets, and produced a long list of candidate issues against standard heuristics.

The principal on the project read the summaries against the funnel, threw out most of the list, merged duplicates, and priced the five that survived.

Published research explains why that order matters. A University of Washington preprint (July 2025) found GPT-4 caught more heuristic issues than a single expert, but also produced 43 to 52 non-issues where experts produced 3 to 6, and caught half or fewer of the problems that run across several screens. F1 is a cross-screen problem. It is the one that matters most here.

Section 5: next-phase scope

The report ends with the work it recommends, in order of return per week of effort:

1. F2 and F3 together, one two-week output, because they are small and fast
2. F1, two two-week outputs, because it changes the first session
3. Re-measure trial-to-paid on the next full month of trials

Each phase has a fixed scope and a fixed price, quoted after the Sprint, not before it. Every two weeks there is one defined output and a recorded walkthrough instead of a status call.

Section 6: the guarantee check

Tallyroom would book the Sprint without the technical architecture review, at the lower end of €5,500 to €8,500.

The Clarity Sprint carries a written guarantee: if it does not surface at least ten times its fee in annual impact, you get the fee back. Time-boxed and fee-capped, full terms in the agreement.

At €6,500, ten times the fee is €65,000. This invented report found $760,320. It clears the bar by a wide margin. A real report would state the result of this check either way.

What a Clarity Sprint report doesn’t contain

No redesigned screens. No promise that conversion will move by a set amount. No findings priced without evidence.

It contains a number, the reasons behind it, and the order to fix things in. If the numbers say the gap is too small to be worth closing, the report says that too.

If your product had a report like this tomorrow, which finding do you think would be at the top?

Let’s find out on your real numbers. Book a free 30-min call, or read the terms on the Clarity Sprint page.

---

Sources

• ChartMogul, The Conversion Report (February 2026)
• Zhong et al., Synthetic Heuristic Evaluation, arXiv (July 2025)
• Nielsen Norman Group, Severity Ratings for Usability Problems