CRO without A/B tests: how small sites find their fixes

Waiting for enough traffic to run an A/B test can take longer than a year. You can find what’s costing you enquiries without one.

Many small sites don’t have enough traffic to A/B test

An A/B test shows half your visitors one version of a page and half another, then compares how many in each group enquire.

On a page where 3% of visitors enquire, reliably spotting a 20% rise (from 3% to 3.6%) takes about 26,100 visitors. That figure comes from Evan Miller’s sample-size calculator at its standard settings, with the effect set to Relative. At 2,000 visitors a month to that page, that’s about 13 months.

Traffic isn’t the only problem. Google’s own testing tool, Optimize, was retired on 30 September 2023, and Google Analytics 4 (GA4) has no built-in replacement. Conversion rate optimisation (CRO), the work of getting more visitors to act, doesn’t have to start with a test.

Read the path in GA4, split by device

Your analytics already shows where people give up, once you split it by device. Before you build either report below, check two settings:

• Key events. Mark each enquiry action, such as a form submission, as a key event: GA4’s name for an action that matters to your business.
• Data retention. GA4 keeps the data behind these reports for 2 months by default. Switch it to 14 months now, so you’ll have older numbers to compare against. The setting only affects explorations and funnel reports, and lowering it later deletes data.

In GA4, build a funnel exploration (a report of how many people get through each step) from the page most visitors land on to your thank-you page. Add Device category as the breakdown, so each step splits by device type, such as mobile and desktop.

Then run a path exploration backwards from your contact page to see what people did just before they reached it. Start with mobile: in 2025, 97% of Australian adults went online using a mobile phone.

Watch session recordings where people drop off

Session recordings (replays of how real visitors moved through a page) show what went wrong at the step that loses people. Microsoft Clarity is free, with no traffic limits. Filter to sessions that reached that step and look for the warning signs Clarity flags:

• Rage clicks: rapid, repeated clicks in one spot.
• Dead clicks: clicks that get no response, which can mean something is broken.
• Quick backs: leaving for another page and coming straight back.
• Click errors: clicks that set off an error in the page’s code.

Clarity masks what people type into input boxes in every mode, and masked content is never uploaded. Recordings still show how each visitor used the page, which raises the same privacy questions as tracking pixels, the small bits of code that report what visitors do.

Australia’s privacy regulator, the OAIC, has advice on tracking pixels. If the Privacy Act covers your business, the OAIC says to collect as little personal information as you can through them, and to say in your privacy policy that you use them. Do the same for session recordings. This is general information, not legal advice.

Have a few people try the main task on a phone

Watching a few people use your site on a phone finds problems your analytics can’t explain. Ask about five people like your customers to do the main task, such as requesting a quote, on their own phone. Watch without helping, and note where they stall.

A 2000 model by Jakob Nielsen suggests five similar users reveal about 85% of usability problems. As the Nielsen Norman Group (NN/g) points out, this kind of test finds what’s broken but can’t measure your conversion rate (the share of visitors who enquire).

Fix clear faults without testing them

Faults that stop people finishing the task, such as a form that rejects a valid phone number, don’t need a test. Fix them. Enquiry forms that get finished on a phone covers each of these common form fixes in detail. Most come from NN/g, and the last one comes from WCAG 2.2, the main web accessibility standard:

• Put labels close to their fields, and don’t rely on placeholder text (the grey hint inside a box).
• Use one column, and cut optional fields.
• Accept common formats, such as spaces in a phone number, and state any format you really need.
• Show error messages that are specific and easy to see.
• Make buttons and links big enough to tap: WCAG 2.2’s minimum is 24 by 24 CSS pixels, with some exceptions.

Write down what each change should do

Before each change, write down the signal that would show it working. HMRC, the UK’s tax office, uses a three-part statement for this. It sets out a guess that you can check after the change. For example:

• We believe that mobile visitors abandon the quote form because the phone field rejects spaces,
• so if we accept phone numbers in any format,
• we will see more mobile form submissions.

Watch a second number too, such as phone calls. The change might only mean that people who would have called you now fill in the form instead.

Record a baseline, then compare before and after

Before you change anything, record a baseline: your enquiry count for a set number of whole weeks, split by where visitors came from (traffic source) and device. Include calls and bookings from your Google Business Profile, which your site’s analytics may never see. The profile’s performance figures only count calls from taps on its call button, and bookings made through a booking provider.

After the change, count the same things for the same number of weeks. A before-and-after comparison is cheap, but UK Government guidance warns it can’t rule out something else causing the change. Note anything else that shifted, like a seasonal rush, and read the result with that in mind.

A/B testing gets faster as traffic grows

At 10,000 visitors a month, the same test takes about 2.6 months instead of 13. A test spots a bigger rise sooner. Taking a 3% enquiry rate to 4.5%, a 50% rise, needs about 4,340 visitors in total. That’s about two weeks at the same traffic.

Last checked 10 October 2026.