Product page CRO

What is actually being bought

The arithmetic behind this service is unusually simple, which is why it is often the easiest thing on a store's list to justify.

An owner running paid traffic has a fixed monthly spend and a mostly fixed cost of goods. The number of visitors is not the lever any more — it is bought and it is priced. The only cheap revenue left is a larger share of the people already arriving. Every extra order that comes from a page working better rather than from more spend goes almost entirely to the bottom line, which is why a page rebuild pays for itself in a way that most marketing work cannot demonstrate.

The second thing being bought is quieter. A product page is where a stranger decides whether to trust an unfamiliar brand with their card details. They arrive with a specific set of questions — will it fit, how big is it really, when will it get here, what does delivery cost, what happens if it is wrong, has anybody else bought this. If the page answers those questions in the order they occur, the purchase happens. If it answers a different set of questions beautifully, the visitor leaves and no one ever finds out why. Nobody fills in a form to say they gave up.

In practice, most of the improvement on most stores comes from three places. The Add to Cart button is not visible where people are actually reading, on a phone. The delivery cost appears for the first time at the final step, after the buyer has invested effort. And the photography never establishes scale, so the customer cannot picture the thing in their hand and postpones the decision permanently.

How the work runs

1. Data first, opinions later. The funnel in GA4 to see where people stop, then session recordings to see what they did before stopping. Recordings are the part that changes minds — watching six people reach the same point and leave is worth more than a page of theory. Two to three days.
2. Buy the product ourselves, on a mid-range phone. All the way through checkout, on mobile data, cold cache, as a first-time visitor. Every friction point written down as it happens — how many fields, at which step the shipping cost first appeared, whether guest checkout exists, where we hesitated. This is where most of the real findings come from. One day.
3. A ranked audit, tied to observed behaviour. Each item states what was seen, what it costs, and what the change is. Ranked by likely impact rather than by ease, with the items we are deliberately leaving alone listed and explained. Two to three days.
4. Implementation. Page order rebuilt for a phone, image sequence chosen so each one answers a specific question, variant selection made obvious, a persistent Add to Cart, delivery cost and date surfaced early, reviews raised, and an FAQ built from the questions your support inbox actually receives. Copy rewritten to replace adjectives with checkable facts. Four to seven days.
5. Thirty days of watching, then a report. Same measurements, same conditions, with a clear line between what moved beyond normal variation and what your traffic volume simply cannot prove either way. Thirty days elapsed, half a day of analysis.

The tools are deliberately unglamorous — Microsoft Clarity for recordings and heatmaps, GA4 and the platform's own analytics for the funnel, PageSpeed Insights where speed is implicated in the drop-off, and Figma when a layout needs approving before it is built.

What moves the price

The range is $400–$2,500 over 1–3 weeks.

How far past the audit you want us to go. A written audit with a ranked list is the smallest version and the cheapest. Audit plus implementation is the middle. Audit, implementation, new copy and a shot list for photography is the top of the range, and it is where most of the durable improvement lives.

How many templates are involved. One flagship product page is a contained job. A catalogue where twenty pages each have their own layout because they were built at different times is a different scope, and the first useful thing to do there is often to make them one template.

How much information you can hand over. Rewriting a page around checkable facts requires the facts — materials, measurements, construction, care, delivery times. Where that exists, the work is fast. Where it has to be extracted from a supplier one email at a time, it is the critical path.

Where this work goes wrong

Taste gets recommended instead of answers. Lighten the palette, change the typeface, move the button. That is an opinion, you have one too, and the discussion goes nowhere. The version that ends the discussion sounds like this — five of the eight people we watched reached the size section and left, there was no size reference anywhere on the page, so a scale image and one line of measurement copy are going in. Nobody argues with the second sentence, and it is also the one that turns out to be right.

The page is only ever judged on a desktop. It is reviewed on a large screen, on fast broadband, with the images already cached, by people who know the product. The majority of the traffic is on a phone, on a mobile network, seeing the brand for the first time. The single most common finding in this work is a call to action that is perfectly visible in the review and entirely below the fold in reality.

A test is run on traffic that can never settle it. A split test is started, one variant is ahead after ten days, it is declared the winner and rolled out. On a store with modest order volume that gap is noise, and the same test run again would as easily favour the other side. Below a certain volume the right method is not testing — it is finding defects, fixing them, and being honest that the evidence is directional rather than statistical.