One number: CVR 2.2% → 4.3%

The account we walked into

On paper, a healthy DTC brand: strong product, loyal customers, a site that converted organic traffic well. Paid search told a different story. Around £11,000 a month bought clicks that converted at 2.2% while the apparel field median sat at 4.50%, meaning the store paid full price for traffic that produced half the field's outcomes. The founder's hypothesis was weak ad copy. The audit's finding was less flattering and cheaper to fix: the account bought the wrong clicks, sent them to the wrong pages, and then undercounted whatever they did anyway.

The undercounting deserves its own sentence. The consent banner, installed in 2023 and never revisited, killed every Google tag the moment a visitor declined, rather than running Consent Mode v2 in its consentless mode, the setup Google has expected of EEA and UK advertisers since March 2024. With a typical ecommerce consent rate in the 45–55% band, that configuration deletes nearly half the evidence before optimization even starts.

Where does a 2.2% conversion rate come from?

Never from one cause. The launch audit split the gap to benchmark into three parts, each with a different owner. Queries the brand should never have bought: generic "cheap trainers" bargain traffic converting near zero, worth 23% of non-brand spend. Right queries landing on wrong pages: category searches sent to a scrolling brand-story page instead of the category grid. And the measurement hole above, which made every prior optimization decision a coin flip. None of these is exotic; all three are what an account looks like after two years without a weekly routine.

What moved the number, week by week?

Weeks four and five are the honest part: the landing-page remap barely moved the headline number at first, because the biggest-volume queries were already routed correctly. It paid off in week six as the long tail caught up. Single-metric reporting makes that visible; blended dashboards would have hidden it under a green arrow.

What we deliberately did not do

No new landing pages, no site redesign, no budget increase, no Performance Max. Each was proposed at some point and declined with a reason in the log: the category pages already outconverted anything we could build in seven weeks, the budget was adequate once it stopped buying bargain hunters, and PMax without 30-plus monthly conversions of clean data would have been bidding on the very noise we had just cleaned. Restraint is a deliverable; it just never gets a slide.

What does 4.3% mean against the benchmark?

The 2026 LocaliQ median for apparel, fashion and jewelry is 4.50% conversion on search, on a $4.44 median CPC. Landing at 4.3% means this store now converts like a competent apparel advertiser, from a starting point in the bottom quartile. We did not beat the market; we removed the reasons the store was losing to it. The remaining gap is a product-page problem, and we said so in the report rather than promising paid search could close it.

What carried over into the routine

The weekly query review that produced the 214 negatives is now standing paid-search routine on the account, and the consent repair pattern became the store's release checklist. The feed groundwork done in week one later fed the Shopping expansion, the same sequence described in the refill-goods write-up. Carbon context: search-only media, estimated at the light end of the 1–5g per impression range per our method, and the removed bargain-query spend was also the account's biggest emissions cut of the quarter, since nothing emits like an impression that converts nobody.