A conversion rate optimisation audit is the diagnostic step that should happen before anybody runs a test, changes a button or rewrites a headline. Done properly it tells you where the money is leaking, whether your site can support testing at all, and which fixes are worth paying for.
What a conversion rate optimisation audit actually is
A conversion rate optimisation audit is a structured review of the route a visitor takes from landing on your site to becoming a customer, built on your own analytics rather than on best practice lists. The output is not a redesign. It is a ranked set of hypotheses, each one tied to a number you can already see in your data, plus an honest verdict on whether your traffic volume can prove any of them.
That last part is the one most audits skip, and it is the reason so many CRO programmes disappoint. If a page gets 400 visits a month and converts at 2%, you are working with eight conversions. No test on that page will reach significance inside a quarter. A real audit says so in writing and points the budget somewhere it can work instead. We cover the wider programme design in our UK CRO playbook, but the audit is where every engagement starts.
It is also worth being clear about what a CRO audit is not. It is not a heuristic walkthrough where somebody scores your site out of ten against a template. It is not an accessibility review, though it will often surface accessibility problems that are costing you money. And it is not a one off deliverable that sits in a folder. An audit that does not end with a prioritised test queue has not finished.
The six areas a CRO audit has to cover
Skip any one of these and the audit produces recommendations that cannot be trusted. The order matters too, because each layer depends on the one before it.
Measurement integrity
Before anything else, confirm the analytics are telling the truth. Duplicate key events, form submissions firing on page load, cross domain tracking broken at the payment step: every one of these invents conversions that never happened.
Traffic quality and segmentation
A site wide conversion rate is an average of things that should never be averaged. Branded search, paid non brand, organic long tail and direct behave nothing like each other, and lumping them together hides both the wins and the losses.
The funnel drop off map
Plot the actual step by step fallout from landing page to thank you page. The biggest single percentage drop is rarely where the team assumes it is, and it is almost never the button colour.
Page level friction
Load speed on real devices, form field count, unexplained costs appearing late, mandatory account creation, trust signals missing at the point of commitment. These are the classic revenue leaks and they are cheap to fix.
Qualitative evidence
Session recordings, on site polls and a handful of customer service transcripts. Numbers tell you where people leave. Only qualitative evidence tells you why, and a hypothesis without a why is a guess.
Test capacity maths
Work out, per page, how long a test would need to run to detect a realistic uplift. This is the gate that decides whether the answer is a testing programme, a straight redesign, or more traffic first.
How to run a CRO audit, step by step
This is the sequence we use. It is deliberately boring at the start, because the interesting findings are worthless if the foundation underneath them is wrong.
- Fix the definition of a conversion. Write down, in one sentence, the single action that makes you money. Enquiry forms, phone calls, completed checkouts. Then check that action is actually recorded as a key event in GA4 and that it fires once, not three times.
- Validate the tracking. Complete the journey yourself on a real phone and a desktop browser, in a private window, and watch the events arrive in the realtime report. You will find something. Everybody does.
- Build the baseline table. Twelve months of sessions, conversions and conversion rate, split by channel and by device. Twelve months rather than three, because seasonality in the UK is severe and a quarter on quarter read can invert the picture entirely.
- Map the funnel. Landing page, category or service page, detail page, basket or enquiry form, checkout or submission, confirmation. Record the percentage surviving each step, per device.
- Find the biggest absolute loss. Not the biggest percentage drop, the biggest number of people lost. A 5% drop on a step 40,000 people reach beats a 40% drop on a step 300 people reach.
- Watch the recordings for that step only. Twenty sessions is usually enough to produce three or four concrete hypotheses. Watching recordings at random is a good way to lose a day.
- Write the hypotheses as testable statements. Each one needs a change, an expected direction, a metric and a reason grounded in the evidence you just collected.
- Rank by expected value and run the sample size maths. Estimated uplift times the traffic reaching that step, divided by implementation effort. Then check each one can actually conclude in a sensible timeframe. Anything that cannot becomes a straight change rather than a test.
The sample size problem, and why it decides everything
Most UK small and medium sized businesses do not have the traffic to run classic A/B testing on anything other than the homepage and the main landing page. That is not a failure, it is arithmetic. Detecting a realistic 10% relative uplift on a 2% baseline conversion rate needs thousands of visitors per variant, and most service businesses do not see that in a month.
The honest response is to split the work into two buckets. Changes that are obviously correct and carry no real risk, such as removing a redundant form field, fixing a mobile layout that pushes the call to action below three scrolls, or adding the delivery cost before the checkout step, should simply be made. You do not need to test whether people prefer being told the postage price up front. Changes that involve a genuine trade off, such as a different pricing structure or a shorter versus longer sales page, need either a test or a before and after measurement with a long enough window and a clear note that it is not a controlled experiment.
Where traffic really is too thin for either, the audit should say that the constraint is acquisition, not conversion, and hand the budget to organic search or paid media until there is something to optimise. An audit that recommends a twelve month testing programme to a site with 900 monthly sessions is selling, not advising.
Five findings we see on almost every UK site we audit
These are not exotic. They are the ones that recur, and between them they usually account for most of the recoverable loss.
- The mobile call to action is below the fold on the devices people actually use. Desktop preview looks fine. On a mid range Android at the screen width most UK traffic arrives on, the primary action is two scrolls down.
- Key events are double counting. A thank you page that is reachable by refresh, or a form plugin firing both its own event and a tag manager trigger, inflates the conversion rate and quietly ruins every decision made from it.
- Delivery or set up costs appear at the last step. This remains the single most reliable cause of abandonment in UK ecommerce, and it is entirely self inflicted.
- The enquiry form asks for information the business does not use. Company size, budget range, how did you hear about us. Every optional field that looks mandatory costs enquiries.
- Trust evidence sits on a separate page nobody visits. Case studies, accreditations and reviews live under About rather than next to the button that asks for money. Design decisions like this are covered in more depth in our guide to how web design affects conversion rate.
What a CRO audit costs in the UK
Pricing across the UK market varies more than it should, largely because the word audit covers everything from a two hour heuristic review to a four week data project. As a rough guide, a genuine data led audit with funnel analysis, session review and a prioritised test queue is a multi day piece of work and is priced accordingly. A free audit is a sales tool, which is fine as long as everybody knows that is what it is. The useful question to ask a supplier is not the price but the deliverable: how many hypotheses, ranked how, with what evidence attached to each, and does the report include the sample size maths that says whether they can be tested.
Peak Digital includes CRO within our paid media retainers and runs it standalone by quote, and every engagement opens with an audit before any test is designed. The detail sits on our conversion rate optimisation service page.
Turning the audit into a programme
The audit is only useful if it changes what happens next week. Three things make that likely. First, a single ranked queue that everybody can see, with the top item already specified well enough to build. Second, a fixed cadence, because testing that happens when somebody remembers produces no compounding at all. Third, a rule for what counts as a result, agreed in advance, so that a flat test is recorded as a flat test rather than quietly re run until it wins.
It also helps to write down what you will not do. Most audits surface twenty or thirty possible changes, and the discipline is in ignoring the bottom twenty. If you want the underlying measurement set up right first, the GOV.UK Service Manual guidance on user research is a good non commercial reference for the qualitative half, and Google's own documentation on key events in GA4 covers the quantitative half.
How a CRO audit differs from a website audit or an SEO audit
The three get confused constantly, and buying the wrong one is expensive. A website audit asks whether the site is technically sound: broken links, page weight, mobile rendering, accessibility, security headers. An SEO audit asks whether search engines can find, understand and rank the pages. A conversion rate optimisation audit asks a narrower and more commercial question: of the people who already arrived, why did most of them leave without doing the thing you need them to do.
They overlap at the edges. A page that takes nine seconds to load on 4G is an SEO problem, a technical problem and a conversion problem at once. But the method is different. An SEO audit is largely deterministic, in that there is usually a correct answer about a canonical tag or a redirect chain. A CRO audit is probabilistic: it produces ranked hypotheses about human behaviour, and the ranking is the deliverable. If a supplier hands you a CRO audit that reads like a list of technical defects, you have been sold a website audit with a different cover page.
The practical sequencing for most UK businesses is to fix the technical and measurement layer first, because a CRO audit built on broken analytics produces confident nonsense, then run the conversion work, then keep both under review. Our website audit checklist covers the technical half if that is the layer you need first. Where an audit keeps returning to the same structural limits, a rebuild is the honest answer, and that is what our ecommerce website development service handles.
FAQ
What is a conversion rate optimisation audit?
It is a structured review of how visitors move through your website, built from your own analytics rather than a best practice checklist. It identifies where people drop out, why they drop out, and which fixes are worth making first. A proper audit also states plainly whether your traffic is high enough to prove any of its recommendations through testing.
How do you do a CRO audit?
Validate the tracking first, then build a twelve month baseline split by channel and device, map the step by step funnel, find the step losing the largest absolute number of people, watch session recordings of that step only, write testable hypotheses, and rank them by expected value against the sample size each would need. Skipping the tracking validation is the most common way an audit produces confident but wrong recommendations.
How much does a CRO audit cost in the UK?
It varies widely because the word audit covers everything from a two hour heuristic review to a multi week data project. Judge suppliers on the deliverable rather than the price: how many hypotheses, what evidence sits behind each one, how they are ranked, and whether the report includes the sample size calculations that say which of them can realistically be tested. Free audits are sales tools, which is fine provided that is understood.
Is a 20 percent conversion rate good?
It depends entirely on what is being counted. Twenty percent is excellent for an ecommerce checkout and unremarkable for a branded search landing page where the visitor already intended to buy. Compare like with like: same channel, same device, same step of the funnel, same definition of a conversion. A site wide average is an average of things that should not be averaged.
How long does a CRO audit take?
A data led audit typically takes one to three weeks of elapsed time. Most of that is not analysis, it is waiting for enough session recordings to accumulate on the specific step under review and checking that tracking fixes have bedded in. An audit delivered in two days has almost certainly skipped the qualitative evidence.
Do I need a CRO audit if my traffic is low?
An audit is still worth doing, but expect the answer to be different. On a low traffic site the audit should identify the obvious friction to fix outright, then say plainly that controlled testing is not viable yet and that the budget belongs in acquisition until there is enough traffic to measure. Anyone proposing a long testing programme on a few hundred sessions a month is selling rather than advising.
