Creative Diversity Audit

Purdy & Figg launched 2,658 Meta ads in the 60 days to 2026-08-18, built from 60 distinct headlines across 46 days on which anything shipped. Price leads 67% of the window, but the mix is moving fast enough that the average is misleading: value and economics falls from 96% of June launches to 46% of August ones, while urgency is absent for two months and then arrives at 26%. Their sharpest arguments, a lab comparison against named competitors and a wall of customer quotes, are baked into the images and appear in no ad copy at all, so a text-only read of this account misses them entirely. Whole lanes stay empty, the refill subscription the product is built around is a single untested execution, and the lane with the best delivery evidence is the one they make least.

Five Things Their Ad Account Says

1. 2,658 ads in 46 launch days, and 60 headlines between them. That is roughly 58 new ads on every day they ship. The volume is not variety: it is one small set of claims permuted across a large creative pool by dynamic delivery, so the account looks enormous and argues a handful of things.
2. The mix is moving fast, and the 60-day average hides it. Value and economics runs at 96%, then 74%, then 46% of launches across June, July, August. Price still leads overall at 67% of the window, but that single number describes a period that is already over. Urgency does not appear at all until August, when it arrives at 26% of the month's launches.
3. The lane with the best delivery evidence is the one they are scaling. Education / how it works reaches Meta's top impression decile on 15% of its ads once age is controlled for, against 8% for value and economics. It is also the only argument rising in every month of the window, from 4% of June launches to 10% of August ones. On a small base, so a signal rather than a verdict, but production and evidence agree here.
4. Format is not the lever. The split is 64% static to 36% video, and both formats reach the top impression decile at effectively the same rate (8.9% video against 8.8% static, a gap of 0.1 points). Whatever separates their winners from their losers, it is not whether the asset moves.
5. Their sharpest claims are in the pictures, where no copy analysis would find them. 57 of the 444 statics we classified by eye carry an argument that appears nowhere in the ad text: a Eurofins lab test run head to head against Zoflora, Method and Smol, side-by-side comparison tables against “supermarket spray” and “big brands”, and a wall of customer quote cards. Sweep all 2,658 ads for any of it and the copy returns zero. Read the text alone and this looks like a pure discount account. It is not.
6. Whole arguments are missing, and one is hiding in plain sight. Gifting / occasion, Press / PR / authority, Founder / craft / origin story return nothing at all, and a raw keyword sweep of every ad confirms those are real absences rather than a classifier miss. The subscription the entire product is built around is not quite a zero, but close enough to be one: it appears in the copy of no ad at all, and in the creative of exactly 1 we looked at, offering skip or cancel any time and 15% off every refill. One execution, unsupported by a single line of text. Separately, problem and solution language appears in 579 ads' body copy but leads not a single headline, so the mess their product cleans up is never the argument, only the supporting detail.

Three Lenses, Because Meta Does Not Publish Competitor Spend

Meta's Ad Library shows which ads are running. It does not show budget, impressions or results for UK commercial ads. Nobody can tell you Purdy & Figg's spend from public data, and any report that puts a pound figure on it is guessing. So this audit uses three separate proxies, kept apart on purpose rather than blended into one score.

What Changed Inside The Window

An account this size averages badly. Splitting the window by launch month shows the mix moving quickly enough that any single blended figure describes a period that has already passed.

Argument Lanes

Every ad mapped onto the same fixed checklist of 14 arguments used for our own client audits. The checklist is fixed rather than derived from what Purdy & Figg happens to run, because a taxonomy built from an account can only describe it. Empty rows are the finding.

Content Type Lanes

The same treatment for the 24 canonical content types. What an ad looks like cannot be read from its copy, so this axis is classified by eye from the creative itself. That needs a still per ad, and pulling stills at the scale of the full window is what triggers Meta's rate limiting, so it is measured on a seeded random sample of 684 ads stratified by format and launch month, giving a worst case of plus or minus 3.2 points at 95% confidence. Shares below are of that sample, not of all 2,658 ads. Arguments, format and rank in every other section are measured on the full window.

Person-Led, UGC And Founder

Who is actually in the ad: whether a brand is buying creator content, fronting its founder, or shipping product-only graphics. Sampled on the same 684 ads as the content types above, for the same reason.

Video, Static And Carousel

An earlier version of this audit withheld the format mix, because the static half of that capture had been cut short by rate limiting and any split would have understated statics by an unknown margin. This capture retrieved between 95% and 99% of Meta's own reported count in every window for both formats, so the mix is now measured rather than estimated.

Seven Checks

The Concepts Holding The Account Up

Concepts, not ads. Recuts of one script with a different creator collapse into a single row under the two-axis rule, which is what stops a wall of near-identical variants reading as variety.

Where The Openings Are

This is a competitor teardown, so the plan below is not what Purdy & Figg should build. It is what their creative says about where they are strong, where they are exposed, and which lanes are sitting open. Every figure is measured on ads launched inside the window rather than on the whole back catalogue, because the back catalogue describes a season that has ended.

Sourcing And Its Limits

Source: Meta Ad Library, public UK listing for Purdy & Figg, captured 2026-08-18. Scope is everything launched between 2026-06-19 and 2026-08-18: Meta's date filter matches on delivery overlap rather than launch, so a 60-day query also returns older ads that merely happen to still be running, and those were filtered out afterwards on each ad's own start date. That makes this a production lens, showing what they are making now, not a delivery lens. Older ads may still be consuming budget and are deliberately out of scope. Active ads only, so creative already tested and switched off is invisible. Ads were harvested by scrolling each page-scoped, impression-sorted window grid to exhaustion, with coverage checked against Meta's own reported count for that window; every window in this capture returned between 95% and 99%. Two guards run throughout: a positive control that must paginate past the first page of results before any harvesting is trusted, and a per-window check that refuses to record a page-one result as a complete one. Start dates come from the start_date field inside each ad's own archive record. Headlines were read from the position the Ad Library card places them in, validated at 92.8% exact against 349 ads for which the full ad detail page was also pulled. Arguments were classified from copy, with the rule that a claim in the headline outranks a price attached to it, so an ad only counts as an offer ad when the offer is the argument. Content type, messenger and person-led come from the creative itself. Those three were classified on a seeded random sample of 684 ads stratified by format and launch month, reproducible from seed 20260820. Meta publishes no spend or impression figures for UK commercial ads, so no pound figure, budget split or results claim appears anywhere in this report. No customer quotes, reviews or performance claims beyond what is visible in Purdy & Figg's own public ad copy are used.