Conversion Rate Optimisation Services: Process and What to Expect

Buying more traffic is the default answer when sales stall, but if a site converts poorly, every extra visitor costs more than it should. Conversion rate optimisation (CRO) is the discipline of finding out why visitors do not buy or enquire, and fixing it in a way you can measure.

The trouble is that "CRO services" can mean anything from a one-off checklist to a rigorous experimentation programme. This guide explains what a proper CRO engagement includes, how the statistics behind A/B testing work in plain terms, and how to tell a capable partner from one that just moves buttons around. For store-specific tactics, see our separate guide to Shopify conversion rate optimisation.

What Does a CRO Engagement Include?

A CRO engagement is a cycle of research, hypotheses, prioritisation, testing and implementation, repeated continuously. The deliverables are a research report, a prioritised test backlog, running experiments, documented results and changes shipped to the live site.

The first month is usually phases 1 to 4. Testing then runs continuously, with research refreshed every few months or after big site changes.

What Research Should Come Before Any Changes?

Good CRO starts by finding where visitors drop off (quantitative data) and why (qualitative data). Without that research, changes are guesses, and most guesses either do nothing or make things worse.

• Analytics: funnel drop-off by step, device, traffic source, country and new versus returning visitors. Look for pages with high traffic and high exit rates, and segments that convert far below average.
• Heatmaps: where people click, tap and scroll. They show ignored calls to action, clicks on non-clickable elements and content nobody reaches.
• Session recordings: watching real sessions reveals hesitation, repeated attempts at forms and confusion. Microsoft Clarity, for example, provides session recordings and heatmaps and describes itself as free; Microsoft says it should not be used on sites targeting under-18s.
• On-site and post-purchase surveys: one-question polls such as "What nearly stopped you buying today?" capture objections in the customer's own words.
• Customer and sales interviews: especially for B2B and high-value products, where the reasons behind decisions rarely appear in analytics.
• Usability testing: asking five or so target users to complete a task while thinking aloud quickly exposes friction the team has stopped noticing.
• Heuristic review: an expert walk-through against clarity, relevance, trust, friction and distraction. Useful, but treat findings as hypotheses, not facts.

How Do You Write a Good CRO Hypothesis?

A CRO hypothesis links an observed problem, a proposed change and an expected effect on a defined metric. Writing it down before testing stops the team from rationalising whatever result appears.

A useful template: "Because we observed [evidence], we believe that [change] for [audience] will cause [effect on primary metric]. We will know when [metric] changes by at least [minimum effect] after [sample size]."

For example: "Because session recordings show mobile visitors scrolling past the delivery information and survey responses mention shipping cost surprises, we believe showing delivery cost and timing beside the add-to-cart button on mobile will increase checkout starts." The evidence, audience and metric are all explicit, so the result teaches something whichever way it goes.

How Are CRO Ideas Prioritised with PIE and ICE?

Prioritisation frameworks score each idea on a few factors so the team tests the most valuable ideas first. PIE and ICE are the most common; both are structured judgement rather than precise measurement.

Whichever model you use, score as a group, write down why each score was given, and revisit scores when new research arrives. The value is in forcing an explicit conversation, not in the decimal places. The PIE and ICE frameworks are summarised in more detail by Growth Method.

What Statistics Do You Need to Understand A/B Testing?

An A/B test splits visitors randomly between the current version and a variant, then checks whether the difference in conversion rate is larger than chance would explain. The essential rules are to set the sample size in advance, run full weeks and avoid stopping as soon as a result looks significant.

Key terms in plain English:

• Statistical significance: how unlikely the observed difference would be if the variant made no real difference. A 95% threshold is common.
• Statistical power: the chance the test detects a real effect of the size you care about. 80% is a common target.
• Minimum detectable effect (MDE): the smallest improvement worth detecting. Smaller effects need far larger samples.
• Sample size: calculated from your baseline conversion rate, MDE, significance and power before the test starts.
• Sample ratio mismatch: when traffic does not split in the planned proportions, a sign of a technical fault that invalidates results.

The most common error is "peeking". Evan Miller's widely cited essay How Not To Run an A/B Test shows that checking results repeatedly and stopping when they cross a significance threshold inflates false positives sharply; in his example, checking after every observation turns a nominal 5% significance level into 26.1%. His advice is to decide the sample size in advance and stick to it, or use sequential or Bayesian methods designed for continuous monitoring.

Also run tests for whole weeks to cover weekday and weekend behaviour, avoid launching during major sales unless that is the context you are testing, and segment results cautiously: slicing a finished test into many segments will find "winners" by chance.

Which Tools Do CRO Teams Use?

A typical stack combines analytics, behaviour tools, a testing platform and a survey tool. Google Optimize, once the default free option, closed on 30 September 2023; Google named AB Tasty, Optimizely and VWO as testing providers it was building integrations with for Google Analytics 4 (GA4).

1. Analytics: GA4 or the platform's own analytics, such as Shopify's reports.
2. Behaviour: heatmaps and session recordings (Microsoft Clarity, Hotjar and similar).
3. Testing: a client-side or server-side A/B testing platform suited to your traffic and tech stack.
4. Research: on-site polls, post-purchase surveys and remote usability testing.
5. Documentation: a shared backlog and learning log, so knowledge survives staff changes.

Behaviour and testing tools set cookies or similar identifiers, so in the UK and EU they must respect your consent banner. That reduces the share of visitors you can observe, which is worth factoring into sample size planning.

How Do You Judge a CRO Partner?

Judge a CRO partner on method, transparency and statistical discipline, not on promised uplifts. Ask to see a real research report, a hypothesis backlog and a test write-up, including a losing test.

Questions worth asking:

• What research will you do before proposing changes, and what will the report contain?
• How do you calculate sample size, and what do you do if a test cannot reach it?
• How do you prioritise ideas, and can we see a scored backlog?
• Who builds and quality-checks variants, and how do you avoid page flicker and broken layouts?
• What do you do with tests that lose or are inconclusive?
• How do winning changes get implemented permanently in our site's code, rather than living forever in the testing tool?

Red flags include guaranteed percentage increases, tests declared after a few days, no mention of sample size, and recommendations copied from generic "best practice" lists. For paid traffic specifically, CRO and ad landing pages overlap heavily; see landing pages for Meta and Google Ads and how creative choices feed them in Meta ads for Shopify stores. Product page structure is covered in our anatomy of a high-converting Shopify product page.

How VEATEX Helps

VEATEX is a Shopify, web and software development agency that also builds Peppol e-invoicing platforms for the GCC, working with clients in the US, UK and the Gulf for over 11 years. Our CRO work combines research and analytics with the development skill to build and ship changes properly, from product pages to complex flows such as the custom subscription builder we built for Joyful Cacao and the pergola configurator we built for Rijpex. We report every test result, including the ones that lose. See our CRO and growth services or get in touch.