Mobile conversion benchmarks from 80 audited stores

We audit the mobile buying path of ecommerce stores and score every store against the same fixed rubric. Every figure on this page is read from the audit database when the page loads, so it reflects the corpus as it stands rather than a snapshot someone typed out.

Which checks fail most often

Share of stores where the check was observed and came back missing or only partly present. A check nobody could observe on a store leaves the denominator rather than counting as a pass, which is why the counts differ between rows.

How each area scores

Average score out of 100, lowest first. Cart and checkout is excluded: reaching it requires adding to a cart, which the audit does not do.

What this sample is, and what it is not

These are the stores we have audited, not a random sample of the market. We selected them, mostly for outreach, so they are not representative of ecommerce as a whole and nothing here should be read as an industry statistic. Each audit is a single visit on a phone, as a first-time visitor who has never heard of the brand, and it looks at the homepage, one category page and one product page. It is not a full site review, and a store can do well on things this audit never looks at. Counts sit next to every percentage so you can weigh them yourself.

How this was measured

Each audit loads the store on a mobile viewport and walks the path a first-time visitor takes: the homepage, then a category page, then a product page. Each step is captured three ways, as the first view, part-scrolled, and the full page. A vision model then scores each area against a fixed rubric of ten checks, returning one of present, partly present, missing, not applicable, or obscured for every check. Score math is deterministic and applied afterwards, so the model judges what it can see and never decides the number.

Checks observed on fewer than ten stores are left out entirely, so nothing here is a percentage computed from a handful of stores.

Last audit in this corpus: 2026-11-02. Figures update as new audits complete. If you quote these, please link back so a reader can see the current numbers and the method above.