Guide
Conversion rate optimization
The full guide: what CRO is, the process and where to start.
Analysis method
CRO analysis: how to find what stops visitors converting
A CRO analysis finds where and why visitors fail to convert on your site, before anyone changes a page. This guide covers the method in order: the numbers first, then the pages, then the people, then a ranked list of what to fix or test.
What is a CRO analysis?
A CRO analysis is a structured review of a website that finds where visitors fail to convert and why. It joins four kinds of evidence: analytics numbers, an expert review of the pages, behavior data such as heatmaps and recordings, and what visitors say. The result is a ranked list of problems and hypotheses, not a redesign.
The analysis is the research half of conversion rate optimization. It comes before any change or test, and it decides which changes are worth making. Skip it and you end up testing button colors on a page whose real problem is a shipping cost nobody sees until the last step. Once the analysis has ranked what to fix, our guide to CRO testing covers which kind of test fits each change.
A good analysis answers three questions in order. Where on the site are visitors leaving in numbers that matter? Why are they leaving there? And which fixes are most likely to help, for the least effort? Everything in the method below serves one of those three questions.
What does a CRO analysis look at?
A CRO analysis looks at a site through five layers: quantitative data, a heuristic review of the pages, behavior data, qualitative research with visitors, and the technical layer that decides whether pages load and read correctly for people and for software. Each layer answers a different question, so a sound analysis uses at least three of them.
| Layer | Question it answers | Typical sources | What you get |
|---|---|---|---|
| Quantitative | Where do visitors leave, and which segments convert worst? | Analytics funnels, landing page reports, device and source breakdowns | The pages and steps worth studying |
| Heuristic review | What on the page breaks known usability and persuasion principles? | An expert walkthrough against a fixed set of criteria | A list of likely friction points, each tied to an element |
| Behavior | What do visitors actually do on the page? | Heatmaps, scroll maps, session recordings, form field data | Evidence for or against each suspected problem |
| Qualitative | Why do visitors hesitate or leave, in their own words? | On-page polls, post-purchase surveys, user tests, support and sales logs | Objections, missing information and the language visitors use |
| Technical | Does the page load, work and read correctly on every device, and for AI agents? | Speed tests, error logs, device testing, the HTML before JavaScript runs | Bugs and blockers to fix straight away, without a test |
How to do a CRO analysis in 8 steps
To do a CRO analysis, start with the goal and the data, then move to the pages and the people. Define the conversion, check the tracking, find the drop-offs, segment them, review the weak pages, watch behavior, ask visitors, and finish with ranked hypotheses. Each step narrows where to look for the next one.
Define the conversion and what it is worth
Name the main conversion (a purchase, a demo request, a signup) and the smaller actions that lead to it, such as add to cart or starting a form. Note what each main conversion is worth, even roughly. Value is what lets you compare a problem on a busy page with a problem on a page that sells more per visit.
Check that the tracking is right
Complete each conversion path yourself on desktop and on a phone, and confirm every event fires once, at the right moment, with the right value. Duplicate purchase events, a form event that fires on page load and payment pages on another domain are common causes of numbers that look precise and are wrong. Our guide to Google Analytics conversion tracking covers the checks in GA4.
Find where visitors drop off
Build a funnel from the first step to the conversion and look for the step that loses the most visitors who clearly meant to act. Also rank landing pages by traffic and conversion rate, so a busy page with a weak rate stands out.
Segment before you trust a total
Split every important number by device, traffic source, new or returning visitors and landing page. An overall rate can hold steady while mobile checkout fails, or fall only because more low-intent traffic arrived. The segment that converts much worse at one step than another segment at the same step points to a specific problem.
Review the weak pages against heuristics
Walk through the pages your data flagged with a fixed set of criteria: is the offer clear, is the next step obvious, is there anything that adds doubt or effort, does it work on a small screen? Write each issue down against the element that causes it.
Watch what visitors actually do
Use heatmaps to see where attention and clicks gather on the flagged pages, and session recordings to watch individual visits through the problem step. Look for rage clicks, repeated attempts at a form field, and visitors who scroll past the answer they need. The heatmap tools guide compares the options.
Ask visitors why
Add a one-question poll on the problem page ("What, if anything, is stopping you from buying today?"), survey recent buyers about what nearly stopped them, and read support tickets and sales call notes for repeated questions. Form analytics shows which fields cause the most errors and abandonment. These are the only sources that tell you what visitors were thinking.
Turn findings into ranked hypotheses
Write each finding as a hypothesis with its evidence, then rank the list by expected impact, confidence in the evidence and effort to fix. Split it into three groups: clear errors to fix now, changes worth testing, and questions that need more research. The next section shows the format.
Which numbers belong in a conversion rate analysis?
A conversion rate analysis needs a small set of numbers, each with a fixed definition: conversion rate, step conversion through the funnel, revenue or leads per visitor, average order value where there is a cart, and form completion where there is a form. Track them per segment and per page, and compare each against its own past.
| Metric | How to calculate it | What it tells you |
|---|---|---|
| Conversion rate | Conversions divided by visitors or sessions, times 100 | How often a visit ends in the goal; the headline number, but blind to value |
| Step conversion | Visitors at a step divided by visitors at the step before | Which step in the path loses people |
| Revenue per visitor | Revenue divided by visitors | Combines conversion rate and order value, so a change that sells less often but bigger still shows up |
| Average order value | Revenue divided by orders | Whether a change moved basket size rather than the number of buyers |
| Form completion rate | Form submissions divided by form starts | How much of the loss happens inside the form itself |
| Landing page conversion | Conversions from sessions that began on a page divided by those sessions | Which entry pages earn their traffic and which waste it |
Leave outside averages out of it. An industry or platform figure for "a good conversion rate" mixes sites with different offers, prices and traffic, so it says little about yours. The useful comparisons are your rate against last period, one segment against another, and your rate against the break-even your costs require. You can work out the rate itself with the conversion rate calculator.
How to turn CRO analysis findings into a prioritized plan
Turn each CRO analysis finding into a hypothesis that names the evidence, the change, the audience and the metric you expect to move. Then score every hypothesis on impact, confidence and effort, and sort the list. Clear errors skip the queue and get fixed; uncertain changes on busy pages become tests; weak evidence goes back to research.
A simple hypothesis format keeps the reasoning visible:
Because we saw [evidence], we believe that [change] for [audience] will [effect]. We will know when [metric] moves.
For example: because recordings show mobile visitors scrolling up and down the product page looking for delivery times, and the exit poll keeps asking about shipping, we believe that showing delivery time next to the price for mobile visitors will raise add to cart rate. The example below shows how a short list might sort. It is illustrative, not data from any site.
| Finding | Evidence | Impact | Confidence | Effort | Next step |
|---|---|---|---|---|---|
| Card form rejects spaces in card numbers | Recordings, error logs | High | High | Low | Fix now |
| Delivery time hidden on mobile product page | Recordings, exit poll | High | Medium | Low | Test |
| Headline does not say what the product does | Heuristic review, survey wording | Medium | Medium | Low | Test |
| Returning visitors convert worse than new ones | Analytics segment only | Unknown | Low | n/a | Research further |
Scoring frameworks such as ICE (impact, confidence, ease) and PIE (potential, importance, ease) are versions of the same idea. The exact scale matters less than scoring every item the same way and writing down why. A hypothesis backed by two independent sources, say a recording pattern and a survey answer, deserves more confidence than one backed by a hunch or a single number.
Which CRO analysis tools do you need?
A CRO analysis needs one tool per layer rather than one tool for everything: an analytics platform for drop-offs, a heatmap and recording tool for behavior, a survey or poll tool for visitor feedback, and form analytics where forms matter. An automated audit tool adds a fast first pass that scores a page against known conversion checks.
- Analytics: GA4 or a product analytics tool, set up with clean conversion and step events.
- Behavior: a tool that records heatmaps, scroll maps and session replays on the pages you flagged.
- Feedback: on-page polls and email surveys, plus the support inbox and sales notes you already have.
- Forms: field-level data on errors, time spent and where people give up.
- Automated audit: a scored review of a page's first screen, calls to action, forms, trust, mobile layout, speed, search basics and readiness for AI agents, as a starting list for the manual work.
The full list, grouped by job, is in our guide to conversion rate optimization tools. Several, GA4 among them, can be used without paying, which is enough to start a first analysis on a small site.
Common CRO analysis mistakes to avoid
The most common CRO analysis mistakes are trusting unchecked tracking, acting on site-wide totals, treating opinions as findings, and jumping to a redesign before the evidence is in. Each one produces a confident plan aimed at the wrong problem, which costs more than doing no analysis at all.
- Skipping the tracking check. A duplicated purchase event or a missing step makes every later conclusion wrong, however careful.
- Reading only the total. Device, source and landing page splits often show a problem the overall rate hides.
- Calling an opinion a finding. "The page looks dated" is not evidence. A finding names an element and the data behind it.
- Using one source only. Analytics alone shows where, never why; a survey alone shows what a few people said. Look for two sources that agree.
- Copying a competitor. Their page converts for their offer and their visitors, and you cannot see their numbers.
- Ignoring small screens. Many problems only appear on a phone, where most reviews are not done.
- Testing everything. A broken form needs a fix, not an experiment, and low traffic pages may never reach a test result.
- Analyzing once. Visitors, traffic sources and the site all change. Repeat the analysis after major changes and at regular intervals.
Audit and rebuilds
Where Convertica fits
A full manual analysis takes time and traffic. The free CRO audit runs the page-level part in minutes: give it a page and it scores it on eight checks, for people and for AI agents, and gives you every finding with its evidence and how to fix it. Use it on the page your analytics flags, or as the first pass before your own research.
If you want the fixes built, Convertica's conversion rate optimization services cover two fixed-price rebuilds: the one page rebuild ($497) has Convertica's team build every fix the audit found on one page, and the full site rebuild ($1,997) does the same across your whole site. Both are on the pricing page. Past results, each with its metric, timeframe and source, are in the CRO case studies.
Keep reading
Related guides
Checklist
CRO checklist
The page-by-page checks to run once the analysis points to a page.
Guide
Conversion funnel optimization
The stages of a funnel and a process to fix the leaks you found.
Short answer
A CRO analysis is a structured review that finds where and why visitors fail to convert. You define the conversion, check the tracking, find drop-offs in analytics, segment by device and source, review the weak pages, watch behavior and ask visitors, then rank every finding as a hypothesis to fix or test.
- Check the tracking before you analyze anything: broken events produce confident, wrong conclusions.
- Analytics shows where visitors leave; heuristics, recordings and surveys show why.
- Segment every number by device, traffic source and landing page before acting on a total.
- End with hypotheses ranked by impact, confidence and effort, split into fix now, test and research further.
- No outside benchmark tells you what your conversion rate should be: compare against your own past, your segments and your break-even.
Questions about CRO analysis
What is a CRO analysis?
A CRO analysis is a structured review of a website that finds where visitors fail to convert and why. It combines analytics data, a page-by-page review against usability principles, behavior data such as heatmaps and recordings, and direct visitor research. The output is a ranked list of problems and hypotheses to fix or test.
How do you do a CRO analysis?
Define the conversion, confirm the tracking is accurate, find where visitors drop off, segment by device and traffic source, review the weak pages against usability principles, watch recordings and heatmaps, ask visitors through surveys, then turn every finding into a hypothesis and rank the list by expected impact, confidence and effort.
How is conversion rate calculated?
Conversion rate is the number of conversions divided by the number of visitors or sessions, multiplied by 100. A page with 4,000 sessions and 80 purchases converts at 2%. Pick users or sessions once and keep it, because switching between them changes the rate without any change in visitor behavior.
What data do you need for a conversion rate analysis?
You need conversion and step events from an analytics tool you trust, broken down by device, traffic source and landing page. Add behavior data from heatmaps and session recordings, form field data if a form is involved, and a few short on-page or post-purchase surveys. Numbers show where; the rest shows why.
How long does a CRO analysis take?
It depends on how many pages and methods you include and how much traffic the site gets. The analytics review can be done quickly once tracking is right, but recordings, heatmaps and surveys need enough visits to show patterns. An automated audit of one page, like Convertica's free audit, runs in minutes.
What is the difference between CRO and SEO?
SEO works to bring more visitors from search engines; CRO works to get more of the visitors you already have to take an action, such as buying or signing up. They meet on the page: a page that answers the searcher's question clearly tends to help both rankings and conversions.
What is a CRO analysis tool?
A CRO analysis tool collects or reviews the data used to find conversion problems. Analytics tools measure drop-offs, heatmap and recording tools show behavior, survey and form tools capture what visitors say and where forms fail, and automated audit tools score a page against conversion checks. Most analyses use several together.
Should I fix problems or A/B test them?
Fix clear errors straight away: broken forms, missing information, slow or failing pages, tracking bugs. Test changes where the right answer is genuinely uncertain and the page has enough traffic to reach a result. Low traffic pages are better served by following research and best practice than by a test that never concludes.
