What the ads cost, stage by stage, and what came back

Stages are periods that tested one thing. They cut across Meta campaigns where a campaign was reused: the Identity test and Leads v2 campaigns run through stages 4, 5 and 6. Rates use the human PostHog entrants of that stage as denominator, except stage 1, which predates on-site tracking and uses Meta pixel counts.

Stage 4's rates are dragged down by the LPV test, which bought 515 entrants who almost never reached screen two. Without them the stage has 119 entrants, an 18.5% lead rate and a 9.2% checkout rate. Stage 1 rates are pixel leads and pixel checkouts over link clicks, since entrants were not tracked yet.

Static site on fiftystrong.app, FiftyStrong brand, US women 48–65, $50 a day, lead objective. Two ads: Says Who (15s) and Workout Programme (70s). The paywall was a fake door at $99 / $45 / $16.

Cheapest leads ever recorded on the account. Only Says Who was carried forward. No revenue was possible: the paywall was not wired to a real checkout, so this stage bought information, not customers. On-site tracking did not exist yet, so these are Meta pixel numbers.

Between stages the site became a Next app, the brand became 50Queen, the domain moved to 50queen.com, pricing became a one-off and an email screen was added. All of it at once, with no spend to isolate any of it. Day one spent $88 in a single hour for 5 leads at $16 each; from day two, broad ages, US/CA/GB/AU and a $2.50 cost cap brought leads back to $0.86.

Leads were cheap again but nobody started a checkout: 3 of 240 paywall views, against 5.7% of leads in stage 1. The new long-form sales page was the cause, with median abandon at 20% scroll. This stage paid to learn that the paywall rebuild had broken the bottom of the funnel.

Sales objective optimising for the pixel Purchase event, one Advantage+ ad set across US/CA/GB/AU and ten Western European countries, 13 ads (12 new creatives plus Says Who). New intro screen, three-screen paywall, two pricing arms, embedded Stripe checkout.

Stopped after two days. Meta had nothing to optimise on: the purchase event fires a couple of times a month against the roughly 50 a week it needs, and the live Stripe webhook was not subscribed to the event that would have sent Purchase to Meta at all. The one trial it produced was cancelled with the note that the moves were too simple.

The Identity test ran the same Chair creative from a 50Queen page and a FiftyStrong page, purchase objective, $50 a day each. On 1 September a landing-page-views set and a lead-objective set (Leads v2) were added so the three objectives could be read side by side. The account hit its $200 cap on 31 August and again on 2 September, which is why the stage ends there.

The only stage that has ever collected money: a $14.99 four-weekly plan on 30 August from the 50Queen arm and a $49 twelve-week plan on 1 September from the FiftyStrong arm, both from the Chair creative. The fiftystrong arm also opened a $19.99 monthly trial on 2 September whose card was later declined. The LPV set is why lead rate reads 3.6%: its 515 entrants were cheap and worthless. The identity question itself stayed unanswered at roughly 25 entrants per arm against the 271 it needed.

The domain switched back to fiftystrong.app and every creative was rebuilt for it. The quiz now opens on the goal question, has a belly branch with its own $19.99 monthly and $59 twelve-week arms, and asks for email before the paywall. Three ads: Says Who and Chair on the lead set, Chair on the purchase set. The cap was reset to $200 and spent in about 13 hours.

Same creative, two objectives: lead ads bought clicks at a third of the price, but 2 of 3 purchase-set checkouts became trials while 0 of 14 lead-set checkouts did. The belly branch carried everything, converting 53% of its paywall views to checkout against none on the other branch. Wallet buttons had been broken for every visitor since 30 August and were fixed on 5 September. Nothing was collected; up to $78.99 lands on 12 September if both trials convert.

Cap reset again. Put up with it (13s), Walking selfie (81s), Says Who, Belly before/after and Chair went live in both ad sets at $50 a day each. Meta reached the cap in 37 hours and has delivered nothing since.

Best lead and checkout rates of any stage, and zero sales: 23 checkouts opened, no card entered, no confirmation failures. Meta gave Walking selfie 63% of the money for the worst unit economics, and Chair, the only creative that has ever sold, got 8 impressions. Put up with it was the clear upstream winner at $1.46 a lead. Five ads in one Advantage+ set is not a test, and the page still had no wallet buttons because fiftystrong.app was not yet a registered Stripe payment domain.

Attribution is by the buyer's first quiz URL in PostHog, joined on the Stripe quiz id. All six came through the Chair creative or the Purchases launch; none of the 43 lead-objective checkouts across all stages has produced a sale.

Lead-objective entrants for the lead row exclude stage 1, whose entrants were never tracked; its 539 clicks and 209 leads are not in that denominator. Refunds to date: none.