Conversion rate optimisation agencies follow the same four-step sequence almost without exception, regardless of what they charge or what they call themselves. The pitch decks vary. The underlying process rarely does.
In order: they audit the page, review whatever behavioural data exists, run structured A/B tests, and report back before the cycle starts again. Each phase feeds the next. Most clients want to skip to the testing. Most clients should focus on the audit first.
They start with an audit
No engagement skips this step. Before any testing plan, before hypotheses, before strategy documents, they read the page - the same way a cold visitor would, applying a structured set of conversion criteria. Is there one clear call to action? Does the headline tell the right reader they're in the right place? Does the page address the obvious objection before asking for a commitment?
This phase is almost entirely judgement-based. A good auditor has seen hundreds of pages and recognises failure patterns fast. A weaker one runs a checklist and calls it done. The output is a findings document - typically annotated, ranked by conversion impact, with specific recommendations attached. Some agencies include rewritten copy as part of the deliverable; many hand back observations only and leave the writing to you.
One to five pages is typical, starting with the highest-traffic entry points. The audit is also the one deliverable you own regardless of what happens next with the engagement.
Then they review your behavioural data
A good CRO agency does not look at the page in isolation. They pull whatever data exists - heatmaps, scroll maps, session recordings, funnel drop-off reports, form abandonment rates - and use it to validate or challenge the audit's findings. If the audit flags the CTA as too far down the page, scroll data either backs that up or it doesn't.
This phase also surfaces things a visual audit misses. A session recording might show visitors consistently dropping at the pricing section - a specific friction point that isn't visible from reading the page alone. The data review is where audit hunches become testable hypotheses. It is also where the engagement starts to require real traffic volume. If your page gets fewer than a few hundred visitors a week, the behavioural data is too thin to draw reliable conclusions from. An agency that doesn't ask about traffic volume before scoping the work is worth pressing on this directly.
They run structured A/B tests
Version A of the headline against version B. The form above the fold versus lower down. Two different CTA phrasings. Each experiment is controlled, tracked, and called when one variant produces a meaningful enough difference to be statistically reliable. The winner becomes the new control, and the next test is built on top of it.
There are real constraints here that agency proposals often understate. A/B testing requires enough traffic to reach statistical significance - typically thousands of visitors per variant, depending on your baseline conversion rate and the size of the change being tested. On lower-traffic pages, a single test can take months to produce a reliable result. A testing programme that sounds fast in a proposal can move very slowly once it is running, and agencies are not always incentivised to tell you that upfront.
One position worth stating plainly: countdown timers are the laziest lever in the CRO playbook. Most visitors recognise artificial urgency and discount it. Any agency that leads with urgency tactics before addressing message clarity is optimising the wrong thing.
They report, iterate, and repeat
At the end of each cycle - usually monthly - the agency produces a report covering what ran, what won, what lost, and what to test next. The hypothesis backlog gets updated and the next round of experiments is scoped. Then it starts again.
This is a retainer model by design. In a mature programme, the ongoing iteration is where value compounds over six to twelve months. The first few months are almost always the most productive - obvious wins get found and fixed quickly. Later improvements are smaller, harder to detect, and require more traffic volume and more sophisticated hypotheses. Plenty of businesses find the ROI drops noticeably after the initial phase. Minimum commitments of three to six months are standard, and for good reason: a single test round rarely tells you much on its own.
What the first 90 days actually looks like
Month one is almost entirely setup. The audit is written, data access is established, the initial hypothesis backlog is built, and a testing tool is configured if one isn't already in place. No test results arrive in month one.
Month two, the first experiments go live. Depending on traffic, results may start to emerge by the end of the month or may need more time to reach significance. Month three is where the first meaningful findings land - and where you get a real sense of whether the programme is moving fast enough to justify the retainer. If tests are running to conclusion and producing usable results, the engagement is working. If tests are still waiting on traffic volume three months in, that's a structural problem worth addressing directly rather than waiting out.
The part that does not require a retainer
The audit - the first step every agency takes - is the most accessible part of the whole process. It doesn't require a testing tool, a data analyst, or a multi-month commitment. It requires someone reading the page cold, against a consistent set of criteria, without the familiarity bias that comes from having built the thing.
Ferguson does exactly that part of the job. Paste a URL and in a few minutes you get a scored report covering the same criteria a CRO agency opens with: headline clarity, CTA structure, objection coverage, trust signals, social proof, and load speed. Every finding is ranked by conversion impact, and the fixes come with copy rewritten in your voice - not just a list of things still to do. If the audit surfaces a handful of fixable issues - and it usually does - fix those first. If you fix them and the page still isn't converting, that's when a testing programme starts to make economic sense, because you've already ruled out the cheap, obvious wins. For a full breakdown of what agencies charge and when the retainer makes sense, the CRO agency cost and timing guide covers that in detail.
What does CRO stand for?
Conversion rate optimisation - the practice of increasing the percentage of visitors who take a desired action on a page, whether that's signing up, purchasing, booking a call, or submitting a form. The 'rate' is simply conversions divided by visitors. CRO agencies specialise in this as a dedicated service, distinct from broader marketing, design, or SEO work.
What is the difference between a CRO agency and a CRO consultant?
An agency is typically a team: an account manager, a strategist, a developer to build test variants, and an analyst to interpret results. A consultant is usually one person doing most or all of that work themselves. Agencies cost more and can move faster on execution; consultants are more flexible and cheaper, but their availability is the bottleneck. For most early-stage businesses, a consultant or a structured audit tool is a more proportionate starting point than a full agency retainer.
How much traffic do you need before a CRO agency makes sense?
The practical floor is a few thousand monthly visitors to the specific page being tested. Below that, individual A/B tests can take months to reach statistical significance, which makes a testing programme uneconomical regardless of what the agency charges. If you're under that threshold, an audit and a targeted set of fixes will move the needle faster than a testing programme - and cost a fraction of a retainer.
How much does a CRO agency cost?
Boutique agencies typically run £2,000 to £8,000 per month, with three to six month minimum commitments standard. Freelance CRO consultants are cheaper - usually £75 to £200 per hour, or a flat fee of £1,500 to £4,000 for a one-off audit and recommendations document. Enterprise CRO firms charge significantly more and typically bundle a dedicated testing tool licence and a larger team into the retainer.