Headline
- One creative has ever produced money: I'm 55 · Chair workout. Both purchases ($49 twelve-week on Sep 1, $14.99 four-weekly on Aug 30) and three of the four trials came from it, all inside the purchase-objective Identity campaign. One of those three was cancelled the day it started; the two Sep 5 belly-arm trials are still live.
- Says Who is the click and lead machine and has never converted. Cheapest clicks on the account ($0.27), cheapest pixel leads ($1.22), 17 humans reached the checkout screen, none paid or started a trial.
- Thirteen creatives never got a real test. Everything in the Aug 28 Purchases launch got under $40 before the campaign was paused. The four creatives built on Sep 4 (Belly before/after, Walking selfie, Put up with it, plus 50queen-branded copies) have $0 of delivery.
- The one non-Chair trial is gone. Overhead · Guided workouts produced a trial on Sep 2 (entered Aug 29) that was cancelled on Sep 3 with feedback that the moves were far too simple and not as described. Read that as a creative-to-product mismatch, not a win for the creative.
- Meta's per-ad purchase column is wrong. It credits both purchases to the 50queen identity ad. Stripe and PostHog, joined on quiz id, put one on each brand ad. The total is right; the split is not.
By creative, all campaigns rolled up
Meta for spend, impressions, clicks and pixel events. PostHog humans for on-site steps (Meta data-centre cities and the LPV test excluded). Stripe for money. Says Who's August validation run predates working PostHog ingest, so its PostHog entrants and leads are understated by roughly 420 and 160.
Spend per creative, USD
Link click-through rate, link clicks ÷ impressions
Creatives under $5 of spend omitted from the rate charts: their rates are noise.
Cost per pixel lead, USD
Checkouts, and how many paid or started a trial
Filled bar: reached the checkout screen. Dark cap: paid or started a trial.
| Creative | Ads | Spend | Impr | Clicks | CTR | CPC | Hold | Pixel leads | CPL | PH entrants | PH leads | Checkouts | Trials | Paid | Revenue |
|---|
Hold is thruplay ÷ impressions. Static images have no hold rate. Says Who spend includes $18.31 in the LPV test, which is excluded from every conversion column.
Says Who against Chair workout, step by step
Same funnel, both creatives, PostHog humans only, all non-LPV campaigns. The two lines run close until the last step, and then only one of them has anyone on it.
People at each step, share of PostHog entrants
Says Who: 1,135 entrants → 257 leads → 17 checkouts → 0. Chair workout: 192 → 43 → 16 → 5. Two-sided Fisher on the last step, 0/17 vs 5/16: p ≈ 0.02.
| Step | Says Who | share | Chair workout | share |
|---|---|---|---|---|
| Entered the quiz | 1,135 | 100% | 192 | 100% |
| Gave an email (lead) | 257 | 22.6% | 43 | 22.4% |
| Reached the checkout screen | 17 | 1.5% | 16 | 8.3% |
| Paid or started a trial | 0 | 0% | 5 | 2.6% |
| Paid | 0 | 0% | 2 | 1.0% |
Lead rate per entrant is identical. Chair workout's entrants are five times more likely to reach checkout, and the ones who do actually pay. The confound is the campaign objective, covered under Reading.
By ad, by campaign
Every ad with delivery. The PostHog column reads entrants → leads → checkouts → trials / paid. Ads created and paused with $0 of spend are listed once per campaign.
| Ad | Status | Spend | Impr | Clicks | CTR | CPC | LPV | Pixel leads | CPL | Pixel purch | PostHog |
|---|
Reading
Click-through
Says Who wins every top-of-funnel measure: 8.6% link CTR against 5.3%, a click at a fifth of the price, 22% hold. In the one clean head-to-head, the same Leads v2 ad set on Sep 4 and 5 with both creatives pointing at fiftystrong.app, Says Who bought 97 clicks for $40 and Chair workout 51 for $43. Says Who is the better ad at getting a woman to tap.
Leads
Per dollar, Says Who again: $1.22 per pixel lead lifetime, $2.37 in the Sep 4 to 5 run, against $2.68 for Chair workout in the same run and $7.08 lifetime. Per click the two are level. PostHog leads over entrants are 21 to 23% for Says Who and 22 to 37% for Chair workout. Chair workout's audience is smaller and dearer and at least as willing to give an email.
Sales
All money to date is Chair workout. Says Who has taken 17 humans to the checkout screen and none of them paid or started a trial. Chair workout took 16 and 5 did. The confound is the campaign objective: every Chair workout conversion came from the purchase-optimised Identity ad set, where Says Who has had $0 of delivery. Its Sep 3 Identity ads were created and paused, and its only purchase-objective run is the $17 carried-over ad on Aug 28. The fair statement is that Says Who under lead optimisation does not produce buyers, and Says Who under purchase optimisation is untested.
Cost per buyer
Chair workout under the purchase objective spent $331 for 2 paid customers and 3 trials: $166 per paid customer, $66 per paid-or-trial, against $63.99 of revenue so far. Two of the four trials on the account were cancelled inside the trial: the Chair workout monthly trial on Sep 2, minutes after it started, and the Overhead trial on Sep 3 with feedback that the moves were too simple. The two Sep 5 belly-arm trials (belly_12week and belly_monthly) from the fs.app Chair ad resolve on Sep 12 and decide whether this creative is anywhere near break-even.
Never tested
Belly · Same outfit before/after, Nothing special · Walking selfie, Put up with it, and the 50queen.com duplicates were created on Sep 4 and paused at $0. Workout Programme 70s had one $59 lead run in August with a 29% hold rate and $1.22 per lead, and was never carried into a purchase campaign.
What to do next
- Raise the spend cap. It is $200 and fully spent. Until then every active ad is off.
- Run Says Who under the purchase objective for real money. $150 or more; the paused Identity · Says Who ad already exists. This is the one open question. If its cheap clicks convert at even half Chair workout's checkout-to-pay rate it wins on cost per buyer by a wide margin. If 0/17 holds, retire it from purchase campaigns and keep it for cheap list-building only.
- Keep Chair workout as the purchase-objective control and stop running it under leads. Leads v2 Chair: $62, 32 leads, 0 trials.
- Give the four Sep 4 creatives a purchase-objective test before judging them. Belly before/after and Walking selfie were built for the identity layer; a leads campaign would only measure their hook.
- Read this again after Sep 12 when the two live trials resolve. Per the standing rule, no funnel test resolves on leads or checkouts.
Sources
Meta ad account act_1761459111652871: /insights?level=ad&date_preset=maximum with actions, action values and video watch fields; ads with creative link URLs (the utm_content on each URL is the join key); account spend_cap and amount_spent; account-level daily insights, last 14 days.
PostHog prod project 559701, SQL over events grouped by quiz_public_id: first-seen utm_campaign, utm_content and brand; distinct quiz screens ≥ 2 for progression; email_captured, paywall_viewed, checkout_started, trial_started, purchase_completed. Excluded: utm_campaign = lpv-test, Meta data-centre cities, Hong Kong.
Stripe live mode: charges and subscriptions created since Aug 7, joined to PostHog on metadata.quizPublicId. Excluded: the Aug 21 e2e charge and Eugene's Aug 29 test trials. Trial cancellations read from canceled_at and cancellation_details on the subscription.
Markdown twin with the same tables: .claude/ads/creative-performance-2026-09-07.md.