Conversion optimisation is sold as a testing discipline. For most businesses, at most traffic levels, testing is the least useful part of it. The useful part is finding and removing friction, which requires observation rather than statistics.
The maths nobody mentions
Detecting a realistic improvement, say from 2% to 2.4%, at reasonable confidence requires thousands of conversions per variant. A site producing forty enquiries a month cannot run that test in a useful timeframe. Running it anyway and stopping when the numbers look good is not measurement; it is reading noise and acting on it.
This is not an argument against improving conversion. It is an argument for a different method at your scale.
What to do instead
Watch people use it. Five real users attempting a real task will reveal more in an hour than a quarter of testing. You are looking for hesitation and re-reading, not opinions.
Read your own enquiries. The questions people ask are the questions the page failed to answer. If three people a week ask whether you work with their sector, that belongs on the page.
Ask the sales team. They hear the objections. Objections handled on the page arrive as better-qualified enquiries.
Look at where forms are abandoned. Field-level drop-off is concrete: if half the people who start abandon at "company size", that field is the problem.
The friction that is nearly always there
- No indication of price, so uncertain buyers leave rather than ask.
- A form that asks for more than the stage warrants.
- Slow loading on mobile, which removes people before they read anything.
- No proof near the point of decision; case studies buried on a separate page.
- Unclear next step, or several competing next steps of equal weight.
When testing is worth it
High-traffic pages, ecommerce category and product pages, and paid landing pages with sustained spend. There, testing headline and offer, in that order, can pay for itself. Even then, test big changes: small variations need enormous samples to distinguish from noise.
Measure the right conversion
Optimising for form submissions can produce more submissions and fewer customers, if the change simply lowered the bar. Track qualified enquiries and, where you can, closed business. A drop in raw volume alongside a rise in qualified conversations is a win that a conversion dashboard will report as a loss.
Frequently asked questions
How long should a test run?
At least two full business cycles, usually two to four weeks, and never stopped early because it looks good.
Is a heatmap useful?
For seeing how far people scroll, yes. Click maps mostly confirm that people click the biggest button.
