Two-week expert and heatmap review of four to six templates, returning a ranked register saying which findings to test and which to ship untested.
About this service
Buy this when you can see something is wrong and cannot afford to find out by testing. A store running 40,000 sessions a month at a 1.8 percent order rate needs about eleven weeks to detect a 10 percent relative improvement at 80 percent power. That is three honest tests a year. This review decides which three are worth spending, and separates them from the changes that should simply ship because the evidence is already one-sided and the downside is bounded.
What we review:
Four to six templates, not the site. Usually category listing, product page, cart, the first checkout step and account creation, chosen after we look at where the money actually leaves. Two of us score each template independently against a frame written for that template, then compare notes. A shared frame produces shared blind spots. Where the two passes disagree, the disagreement goes in the deliverable rather than being resolved into a consensus that hides it.
What heatmaps are good for, and what they are not:
Click maps tell you whether an element was found. They say nothing about why it was skipped, and a hot area is often just a large one. Dead clicks on non-interactive elements and rage clicks are the two signals we act on directly, because each names a specific broken expectation. Scroll maps mostly describe your device mix, so we normalise by viewport class before drawing any conclusion about fold position. Zone-level engagement from Contentsquare, or area maps in Clarity where Clarity is what you have, is quoted with its segment and session count attached or not quoted at all.
Consent coverage:
Client-side tools in Sweden, Norway and Denmark typically see 55 to 75 percent of sessions after the consent gate, and the missing quarter is not random. It skews toward first-time visitors on paid traffic, which is the group most reviews are actually about. Every figure we hand you carries its coverage rate. Where a tool cannot report its own consent loss, we do not use its numbers.
What you get:
A findings register. Each entry carries the observation, the evidence with segment and sample, the decision it implies, the size of effect a test would have to detect for the test to be worth its weeks, and our call: test it, ship it, or leave it. Ordered by expected value in euros. Not by an ICE score, which is an opinion with arithmetic on top.
Not included:
No implementation, no design files, no copy rewriting. No tag manager work beyond confirming your recording tool fires where we need it. We do not audit page performance here. If largest contentful paint on the product page is the problem, that is an engineering ticket and you get one line saying so, not four pages of padding.
Who should not buy this:
Teams who have already decided and want a document to carry into a meeting. Flows under roughly 8,000 sessions a month, where nothing we find can be validated afterwards and the register becomes a wish list. Anyone who wants a score out of a hundred. We do not produce one, and the shops that do are selling comparability rather than judgement.
How it runs:
Kickoff, access, two working weeks, then a session with the people who will build the changes rather than a presentation to their directors. The register arrives as a spreadsheet you own and can edit. We do not lock findings inside a PDF so that changing them requires calling us.