Pricing Page Work for Fee Tables and Tiered Plans

Before anything about layout: on a page selling to consumers in France, prices are displayed inclusive of tax, and any discount must reference the lowest price applied in the previous 30 days. That reference-price rule has been in force since the Omnibus directive was transposed in May 2022, and roughly a third of the pricing pages I am asked to test break it before they underperform. That gets fixed first and it is not billable optimisation work. Then the harder part. A pricing page is the worst page on your site to A/B test, and most shops test it as though it were a hero image. Conversion rate is the wrong metric: a variant that sells more of the cheapest plan wins on conversion and loses on revenue. Revenue per visitor is the right metric and it has a long tail, so the usual t-test on it behaves badly; I read it with a trimmed mean or a bootstrapped interval, and I state up front how large an effect your traffic can resolve. On most pricing pages that number is 8 to 12 percent, not 2 percent. If a previous agency reported a 3 percent revenue win, they reported noise. Price levels versus price presentation: These are different jobs and only one is mine to test freely. Presentation is testable: how a fee table is ordered, what a plan is called, which comparison sits adjacent to which, what the annual toggle does, where the fine print lives. Price levels shown simultaneously to the same market are not something I will split-test. Customers talk to each other, the screenshots reach Reddit, and in several markets this is a legal question rather than a matter of taste. Where a price level genuinely has to be tested, it runs on new customers only, cohort by cohort, with the change disclosed. Fee tables specifically: For crypto and trading products the pricing page is a fee schedule, and the number the customer actually pays is not on it. Maker and taker tiers, spread, network fees and the withdrawal charge combine into an effective cost that your best prospects compute in a spreadsheet and your worst discover after their first trade. Publishing that effective cost, for a realistic order size, is the change that most often moves signups and 90-day retention together. It is also the change your growth team will argue against hardest. Guardrails, which are the reason to hire me rather than a designer: Every pricing test carries a 90-day retention guardrail and a refund or chargeback guardrail, both agreed before launch. A pricing win at checkout that costs you at renewal is the most common way a conversion programme destroys value while reporting success, and it is invisible on any dashboard that stops at the transaction. Not included: packaging economics, willingness-to-pay research, discount policy, and any decoy tier whose purpose is to make a comparison misleading. I will design a tier that is honestly worse than the one beside it. I will not design one intended never to be bought. Not for you if: your checkout is leaking and unfixed. The pricing page is upstream of everything and downstream of nothing, and testing it while the payment step drops a fifth of attempts wastes the sample. Also not for teams below roughly 2,000 pricing-page sessions a week, where no test concludes and the work should be qualitative instead. What you get: the display fixes, a rewritten page shipped by your own team, the effective-cost model behind any fee display, a pre-registered analysis plan naming the guardrails, and a read-out that includes the tests that failed.