The site is built well and measured badly.
282 findings across three web properties, three tag containers and three ad accounts. The engineering scores a B. Everything failing is configuration, and most of it is reversible in an afternoon by someone with account access and no engineering ticket at all.
What the platforms are being paid, and what they can see
Every figure below was read from the live account. Move the slider to model what the questionnaire conversions are costing in misdirected optimisation.
Optimising against fiction
Paid to fly blind
$131,568 of that, 71% of all Meta spend, sits on one campaign that reports no result at all: its ad sets optimise for different events, so Meta declines to produce a campaign-level number. No campaign in either account has a computable ROAS.
All 282 findings
Search across evidence text and file paths. Every entry carries its evidence, root cause, impact and fix.
Remediation plan
52 work packages, grouped by who does them. Tick one off and the progress below updates, including how many findings you have closed. Progress is saved in this browser.
Well built. Not shippable.
Two reviewers reached NO-GO independently, from opposite directions: one measured page parity, the other whether the cutover plan is executable and whether the result would function as a site.
The revamp built pages. It did not build a site.
What is missing is the connective infrastructure: canonical tags, sitemap entries, redirects, a 404 page, analytics wiring and a working conversion path. Every v3 treatment page, 111 of 112, ships a primary button pointing at a file that does not exist. None of this is hard. None of it has been done. Do not cut over before fixing measurement, or the defect follows the site across and destroys the before-and-after comparison that would prove the work landed.