Conversion rate optimisation grounded in research

Research first: the four sources that tell you why people leave

Every weak conversion programme we have seen skipped straight to ideas. Someone suggests a change, someone else disagrees, the loudest voice wins, and the result is unmeasurable. Research exists to remove that argument by replacing it with observation.

Four sources answer different halves of the question. Funnel data tells you where people leave and what each drop-off costs, but never why. Session recordings and heatmaps show the behaviour behind the number: rage clicks on something that is not a button, forms abandoned at one specific field, people scrolling past the answer they came for. On-site surveys ask the visitor directly, and a single well-placed question on an exit-intent trigger often produces more insight than a month of dashboards.

The fourth source is the one most agencies never touch: your own sales and support conversations. The objections your team answers on every call are the objections your pages are failing to handle. Reading a month of enquiry threads is usually the cheapest research available and it costs nothing but attention.

• Funnel and path analysis: where the losses are and what they are worth
• Session recordings and heatmaps: the behaviour behind the drop-off
• On-page surveys: the visitor’s own reason, in their words
• Sales and support transcripts: the objections your pages never answer

How we prioritise, and why the biggest leak is rarely the homepage

Once research produces a list of problems, the ordering decides the return. We score each issue on the revenue at stake, the confidence that the fix helps, and the effort to implement. High value, high confidence and low effort go first, which is unglamorous and reliably profitable.

Revenue at stake is calculated from the funnel, not from intuition, and this is where expectations usually change. A homepage redesign feels significant because everyone sees it. But if the real loss is happening on a checkout step or a form that a smaller number of much higher-intent visitors abandon, fixing that step is worth more, costs far less and takes days rather than months.

This ordering is also what makes the programme survivable. Early, cheap, provable wins buy the credibility and the budget for the larger changes later. Programmes that open with a six-month redesign tend to be cancelled before anything is measured.

The friction audit: what actually costs enquiries

Most lost conversions are not caused by persuasion failures. They are caused by friction: small obstacles that individually seem trivial and collectively add up to a decision to leave. Friction is the cheapest thing to fix and the most consistently ignored.

Forms are the usual offender. Every field costs completions, and fields that feel intrusive relative to the stage cost disproportionately. Asking for a phone number before the person knows your price, or for a company size before they know whether you serve them, moves the exchange in the wrong order. Asking for what you genuinely need to have the first useful conversation and nothing else is almost always a net gain.

The other recurring costs are unanswered questions and unclear next steps. If a visitor cannot tell whether you work with businesses like theirs, roughly what an engagement involves, or what happens after they submit, they postpone. Postponement reads as a bounce. Stating the unglamorous specifics, including what you do not do, converts better than another round of adjectives.

• Form fields that ask for more than the first conversation needs
• Prices or ranges absent with no explanation of why
• No indication of who the service is and is not for
• Ambiguity about what happens after the button is pressed
• Pages that load slowly enough on mobile data to lose the impatient

Measuring when you cannot run an A/B test

Formal testing needs volume, and most businesses do not have it on most pages. That does not mean improvement cannot be measured, it means the proof method changes and the honesty about it has to increase.

For lower-volume pages we use before-and-after measurement over comparable periods, with the change isolated so nothing else ships at the same time, and with obvious confounders written down in advance: seasonality, a campaign launch, a price change, a press mention. If a confounder lands mid-measurement, we say so rather than claiming the result.

We also lean on directional evidence that does not need statistics. If recordings showed people abandoning at a specific field and, after the field is removed, abandonment at that point stops, the mechanism is clear even without a significance figure. What we will not do is present a small before-and-after difference as a proven percentage uplift, because that number would not survive contact with the next quarter.

What conversion work cannot fix

Conversion optimisation moves the rate at which interested visitors act. It cannot make an uninterested visitor interested, and this is the honest limit of the discipline.

If traffic is poorly matched, because ads target broadly or rankings come from informational searches with no buying intent, the conversion rate is reporting a traffic problem. Rewriting the page will not fix it, and the correct answer is upstream, in targeting or keyword selection. We say so rather than selling a testing programme against it.

The same applies to the offer itself. If the price is wrong for the market, the guarantee is weaker than every competitor, or the service is genuinely undifferentiated, no amount of layout work compensates. Those are business decisions and we will raise them, but they are yours to make. Conversion work is a lever on an existing proposition, not a substitute for one.