A CRO audit is a structured pass over everything between a visitor arriving and a visitor converting, scored so the fixes come out in priority order. Most published CRO audits give you five to ten steps. This one gives you forty specific checks, each with a threshold you can actually pass or fail, in a spreadsheet you can download and work through.
The difference matters more than the count. “Improve page speed” is not a check, because two people can disagree about whether you passed it. “Largest Contentful Paint under 2.5 seconds at the 75th percentile on mobile” is a check. Everything below is written to that standard, and it is the same instrument we use when we run conversion rate optimization work for clients.
What a CRO audit actually is
A CRO audit is diagnostic, not creative. You are not redesigning anything. You are walking the converting path, recording what is true against a standard, and ending with a ranked list of what to fix first. The output is a scored sheet, not a slide deck.
It is also not the same thing as deciding you need one. If you are still at the stage of working out whether your funnel has a problem at all, our piece on the 7 signs your funnel needs a CRO audit is the better starting point. This article assumes you have decided, and want the instrument.
Sources and further reading
The thresholds quoted here come from primary documentation, read directly rather than from a summary:
Google web.dev: Web Vitals, the Core Web Vitals metrics and their thresholds.
Google web.dev: Defining the Core Web Vitals metrics thresholds.
Google web.dev: Best practices for cookie notices.
Google Search Console Help: the Core Web Vitals report.
Chrome User Experience Report documentation, on field versus lab measurement.
To ensure you’re hitting the recommended target for these metrics for most of your users, a good threshold to measure is the 75th percentile of page loads, segmented across mobile and desktop devices.
AuthorGoogle web.dev, Web Vitals
That sentence is the one most CRO articles skip, and skipping it makes the numbers useless. A threshold without a percentile is not a threshold. If you measure your average load time you will pass comfortably and still be losing the quarter of your traffic having the worst experience, which is disproportionately the mobile traffic on older devices.
The thresholds worth writing down
Three of the forty checks have official numbers attached. Use them exactly as published, at the 75th percentile, measured on field data from real users rather than a lab run.
Metric
What it measures
Passing threshold
How to read it
Largest Contentful Paint
Loading
Under 2.5 seconds
Field data at the 75th percentile, split by device
Interaction to Next Paint
Interactivity
200 milliseconds or less
Cannot be measured in a lab tool with no user, so use field data
Cumulative Layout Shift
Visual stability
0.1 or less
Watch for shifts that happen after the CTA is already visible
Everything else in the checklist is a pass or fail judgment, but each one is written so that two people auditing the same page would reach the same verdict. That is the whole design goal.
What the 40 checks cover
The checklist is split into five sections of eight. The split is deliberate: each section catches a different class of problem, and in our experience teams are usually strong in two of them and have never looked at a third.
Section
What it catches
The failure we see most
Technical and speed
People who leave before the page finishes arguing its case
Third-party script bloat nobody owns
Trust and credibility
People who understood the offer and did not believe it
Proof on a separate page the buyer never visits
Form friction
People who decided to convert and then gave up
Required fields no downstream process uses
Mobile
Problems that exist only on the device most of your traffic uses
Sticky elements covering the CTA on small viewports
Analytics instrumentation
Conclusions drawn from numbers that were wrong
Conversion events firing twice
Download the sheet below, then work through the sections in the order that matches your suspicion. If you have no suspicion, start with analytics, because every other finding depends on the numbers being real.
Eight checks, three of which are the Core Web Vitals above. The other five are the ones that do not show up in a performance score at all: broken assets anywhere on a converting path, a consent banner that blocks the first meaningful interaction, error pages that dead-end instead of offering a route back, site search that returns nothing useful, and a third-party script inventory where every tag has a named owner.
The script audit is the one that pays most often. Tags accumulate because adding one is a five-minute job and removing one requires knowing who asked for it. If nobody can name the owner of a script, that is the finding.
Trust and credibility
These eight are about the gap between understanding your offer and believing it. The checks cover proof placement within one screen of the primary call to action, pricing expectation set before the form rather than after, a contact route that is not a form, findable company identity, attributed rather than anonymous reviews, security marks at the point of payment rather than in the footer, terms reachable in one click from the decision point, and claims that carry a source.
The last one is worth dwelling on, because it cuts both ways. Unsourced result claims do not just fail to persuade the sceptical buyer, they actively cost you credibility with exactly the buyer who was going to scrutinize the contract.
Form friction
Eight checks on the step where people who already decided to convert still do not. Two are worth calling out. First, every required field has to be justified by a downstream process that actually uses it: fields get added over years and almost never removed, and each one costs completions. Second, the data a user typed must survive a validation error. Clearing a form on a failed submit is the single fastest way to lose a lead who had already done the work.
The section also covers field-level abandonment instrumentation, inline plain-language errors, correct input types so mobile keyboards and autofill behave, a clean console on submit, a thank-you state that says what happens next, and a measured first response time. A fast form feeding a slow follow-up still loses the deal.
Mobile
Eight checks that only exist because desktop-first habits persist. The primary action visible without scrolling on a mid-range phone, tap targets at least 44 by 44 pixels, body copy at least 16 pixels, sticky elements that do not cover the action, mobile conversion tracked as a ratio to desktop rather than in isolation, an end-to-end completion on a real device, mobile-sized image assets, and interstitials that dismiss one-handed.
The ratio point is the one teams miss. An absolute mobile conversion rate tells you very little. The ratio of mobile to desktop, trended over time, tells you whether the gap is closing or widening, which is the actual question.
Analytics instrumentation
This is the section almost nobody publishes, and it is the one we would run first. Of the pages currently ranking for this topic, most stop at “set up goals in your analytics and enable event tracking.” That is not instrumentation, it is intent.
The eight checks: each conversion event fires exactly once per completion, verified in a debug view on a real submission. Client-side and server-side events deduplicated by a shared event ID if you run both. Consent state recorded alongside every event, so you can state what share of your data is missing rather than guessing. Bot and internal traffic excluded by a filter that has been verified as actually working. Analytics conversions reconciled with your CRM or order system to within 5% for the same period. Macro and micro conversions reported separately rather than summed. Traffic segmented by new versus returning before any conclusion is drawn. And a dated change log, so a metric shift can be matched to a release.
If the reconciliation check fails, stop the audit. Every other finding you make is being measured with an instrument you have just proved is wrong. This is also why the measurement layer tends to be where the real problem lives when a funnel looks inexplicably broken, a theme we cover in why your user journey is not converting.
How to score and sequence the findings
A list of forty findings is not a plan. The sheet includes a scoring model so the order comes out of the data rather than out of whoever spoke last in the meeting.
Mark each check Pass, Fail or Not applicable. Resist the urge to invent a middle option. Ambiguity is what the thresholds exist to remove.
Score every failed check for Impact, 1 to 5. How much revenue sits behind this specific path.
Score every failed check for Effort, 1 to 5. Be honest about engineering time, not optimistic.
Let the sheet compute Priority. Impact multiplied by (6 minus Effort), so high impact and low effort rises.
Sort descending and draw a line after what this quarter can genuinely absorb. Everything below the line is a backlog, not a failure.
Assign an owner to every item above the line. An unowned finding is a note.
The point of the scoring is not precision, it is removing the loudest voice from the sequencing decision. Two people scoring independently and then comparing is more useful than one person scoring carefully.
What the audit does not tell you
The checklist finds problems. It does not prove that your fix worked, and treating a completed audit as a set of guaranteed wins is the most common way teams waste the work.
Two of the five sections produce findings that are true on their face. If a required field has no downstream consumer, that is a fact, and removing it does not need a test. If conversion events fire twice, that is a defect, and fixing it is not an experiment. Act on those directly.
The rest are hypotheses. Moving proof above the fold, rewriting a thank-you state, changing a CTA: each is a reasonable bet, and each needs validating against the thing that actually matters. Three rules make that honest. Decide the sample size and the minimum effect you care about detecting before the test starts, not after you look. Set the duration in whole weeks so you capture the same mix of weekdays and weekends. And do not stop the moment the line crosses, because a test watched continuously will eventually show you a result you like whether or not one exists.
If your traffic cannot support a readable test on a given page, that is a finding too. It means the page should be improved on judgment and monitored, rather than tested badly and declared a win.
How often to run it
Quarterly for the full forty, and the analytics section again after any release that touches tracking, consent, or the checkout. The analytics section is cheap to rerun and is the one most likely to silently break, usually because a tag was added by someone who did not know the dedup rule existed.
Rerunning the full sheet also gives you something a one-off audit cannot: a dated record of what you checked and what you found, so next quarter’s shift can be matched to a cause instead of a theory.
Where to start
Download the sheet, run the analytics section first, and see whether your numbers reconcile. That single check decides whether the rest of the audit is measurement or guesswork. If you would rather have a second pair of eyes on the result, or you run the sheet and want to know what we would prioritize, get in touch and we will go through your scored sheet with you.
Frequently asked questions about CRO audits
What is a CRO audit?
A CRO audit is a structured review of everything between a visitor arriving and a visitor converting, measured against defined thresholds and scored so that fixes come out in priority order. It is diagnostic rather than creative: the output is a ranked list of findings, not a redesign.
How long does a CRO audit take?
The forty checks take roughly a day of focused work for a single converting path if your analytics access is already in place. The analytics section alone takes an hour or two and is worth running first, because it determines whether the rest of the findings are trustworthy.
What should a CRO audit checklist include?
Five areas: technical and speed, trust and credibility signals, form friction, mobile-specific checks, and analytics instrumentation. Each item needs a threshold or a pass condition specific enough that two people auditing the same page would reach the same verdict.
What are the Core Web Vitals thresholds for conversion pages?
Largest Contentful Paint under 2.5 seconds, Interaction to Next Paint at 200 milliseconds or less, and Cumulative Layout Shift at 0.1 or less. Google specifies measuring these at the 75th percentile of page loads, segmented across mobile and desktop.
How do you prioritize CRO audit findings?
Score each failed check for impact and effort on a 1 to 5 scale, then rank by impact multiplied by (6 minus effort) so high-impact low-effort work rises. Draw a line after what the quarter can absorb and assign an owner to everything above it.
How often should you run a CRO audit?
Quarterly for the full checklist, and rerun the analytics section after any release that touches tracking, consent or the checkout. Tracking is the part most likely to break silently between audits.
Is a CRO audit the same as a UX audit?
No. A UX audit evaluates the whole experience against usability principles. A CRO audit is narrower and more measurable: it looks only at the converting path and scores each finding by its effect on conversion, which is why it ends in a prioritized fix list rather than recommendations.
Send us your scored sheet, or just the failed rows. We will tell you which findings we would act on first, which ones need a test rather than a fix, and what we would check that the forty points do not cover for your particular funnel.