FiftyStrong · Meta · validation test
$106.85 has bought 95 completed quizzes at $1.12 each — three to seven times cheaper than the plan assumed. One of the two creatives is doing most of the work, and the largest remaining leak is not in the ads at all.
Everything below is lifetime-to-date for the only campaign that has ever run on this account.
Spent
$106.85
2.4 days live · using only 81% of the $50/day budget
Quiz completions
95
The Lead event — she answered every question and hit the paywall
Cost per completion
$1.12
3–7× better than the $3.33–$7.50 the copy spec assumed
Paywall taps
8
$13.36 each — a painted door, so read it as intent, not revenue
The read: demand is not the problem. Women 48–65 in the US click these ads at 8.6% of impressions and finish a 25-question quiz three times out of four. Both of those numbers are well above what the plan budgeted for, and they held steady across three days without any audience fatigue — frequency is still 1.24.
Two things are worth acting on. Ad A beats Ad B by 67% on cost per lead, and the gap is statistically real rather than noise. And exactly half the women who land on the quiz never tap a single answer — a bigger pool of lost leads than anything the creative test can recover.
What the test still cannot tell you is whether this is a business. Eight paywall taps against a door that does not open is an intent signal, not a conversion rate. That number needs roughly another $650 of spend before it means anything.
Five pixel events fire between the impression and the paywall. Bars are on a log-ish visual scale so the small steps stay readable; the percentages are the honest step-to-step rates.
Landing page views exceed outbound clicks slightly because Meta counts the two differently — it is not a data error, and the practical read is that essentially every click arrives.
| Step | Count | Rate from prior | Cost each |
|---|---|---|---|
| Impressions | 2,813 | — | $37.98 CPM |
| Outbound clicks | 241 | 8.57% | $0.44 |
| Landing page views | 253 | 105% | $0.42 |
| Quiz starts (Q1 answered) | 126 | 49.8% | $0.85 |
| Quiz completions (Lead) | 95 | 75.4% | $1.12 |
| Paywall taps (InitiateCheckout) | 8 | 8.4% | $13.36 |
| Purchases | 0 | — | no payment form exists yet |
The quiz itself is in good shape. Three out of four women who answer the first question walk all the way through roughly 25 screens to the paywall — that is a strong completion rate for a quiz this long, and it means the questions, the pacing and the loader fix are all holding. The problem sits one step earlier.
Both ads ran in the same ad set against the same audience, so Meta allocated between them on live performance. It gave A 59% of the budget.
Outbound click-through rate
Cost per quiz completion — lower is better
Landing → completion, once she is actually on the quiz
Video watched to 100%
The third pair is the one that changes the decision. Once a woman is on the quiz, B's traffic converts slightly better than A's. B is not attracting worse people — it is simply getting far fewer of them to click. That is a hook problem, not an audience or offer problem.
| Metric | A · Says Who · 15s | B · Workout Programme · 70s | Gap |
|---|---|---|---|
| Spend | $63.00 | $43.85 | Meta chose 59 / 41 |
| Impressions | 1,671 | 1,142 | — |
| CPM | $37.70 | $38.40 | same auction price |
| Outbound CTR | 10.17% | 6.30% | +61% |
| Landing page views | 182 | 71 | — |
| Quiz starts | 91 | 35 | — |
| Quiz completions | 67 | 28 | — |
| Cost per completion | $0.94 | $1.57 | B costs 67% more |
| Landing → completion | 36.8% | 39.4% | B marginally ahead |
| Paywall taps | 7 | 1 | too few to trust |
| Average watch time | 10s of 15s | 17s of 70s | — |
| ThruPlays | 467 | 344 | — |
Is the gap real? Yes, on both measures. The click-through difference is far outside noise at this sample (z ≈ 3.6, p < 0.001), and even the harder test — completions per impression, which is what you actually buy — clears the bar (z ≈ 2.25, p ≈ 0.02). The paywall-tap split of 7 vs 1 does not clear it and should be ignored for now.
Targeting is deliberately broad — US, women, 48–65, no interests, Advantage audience off. So the age and placement splits below are Meta's choices, not yours.
| Age | Spend | Share | CTR | Completions | Cost each |
|---|---|---|---|---|---|
| 45–54 | $12.20 | 11% | 9.88% | 6 | $2.03 |
| 55–64 | $50.62 | 47% | 15.84% | 49 | $1.03 |
| 65+ | $44.10 | 41% | 16.31% | 40 | $1.10 |
| All | $106.85 | 100% | 15.18% | 95 | $1.12 |
The buyer is 55 and up. The youngest band costs roughly double per lead, and Meta has already routed 88% of spend away from it without being told to — which is the argument for not touching the age settings. It is an argument for writing copy to a 60-year-old rather than a 49-year-old.
| Placement | Spend | Share | CTR | Completions | Cost each |
|---|---|---|---|---|---|
| $87.34 | 82% | 15.90% | 79 | $1.11 | |
| $19.27 | 18% | 12.10% | 16 | $1.20 |
An 8% cost difference on 18% of spend. Not worth a placement exclusion — leave both on and let the auction keep the option open.
Three days, only one of them complete. August 8 is the only full, fully-attributed day — and it is the best one.
August 7 is B's worst day and the campaign's most expensive — that is the learning phase, and it corrected within 24 hours. August 9's uptick is largely an artefact: with a 7-day click window, conversions from today's impressions are still landing, so today's cost per lead always looks worse than it will settle at.
| Day | Spend | Impressions | Clicks | Completions | Cost each |
|---|---|---|---|---|---|
| Aug 7 · from 06:16 | $41.17 | 1,091 | 143 | 27 | $1.52 |
| Aug 8 · full day | $40.52 | 1,018 | 184 | 49 | $0.83 |
| Aug 9 · to ~14:50 | $25.18 | 706 | 100 | 19 | $1.33 |
| Total | $106.85 | 2,813 | 427 | 95 | $1.12 |
Note the delivery, not just the cost: a $50 daily budget spending $41 and $40.52 on its two full days. The ad set is under-delivering by about a fifth, with frequency at 1.24 — so this is a cold pixel finding its footing, not an audience running out.
Ordered by expected effect on cost per lead, largest first.
253 women landed on the quiz. 126 tapped an answer. The other 127 left without touching it — more lost leads than the entire A-vs-B gap is worth. Moving that first-screen engagement from 50% to 65% takes cost per lead from $1.12 to roughly $0.86 on the same spend.
Screen one is "Fitness that works with your body, not against it — where are you right now?" with four image cards. Worth testing: fewer cards, a lighter first question, or matching the screen's promise more directly to the line she just clicked in the ad.
B costs 67% more per lead and the difference is statistically real. Pause it and hand the budget to A.
Do not bin the asset. B's traffic converts better than A's once it lands, and it holds 17 seconds of attention against A's 10. The failure is entirely in the first three seconds. Recut B's opening against A's hook and run it again.
You are 3–7× better than the plan's cost assumption, frequency is 1.24, and the ad set is only spending 81% of the $50 it already has. Broad US women 48–65 is nowhere near exhausted.
Double it, hold everything else fixed, and read cost per lead after 48 hours. If it holds near $1.10 you have a scalable top of funnel and the question moves permanently to the paywall.
One winning ad and one loser is not a creative programme. A is a 15-second hook that works — the useful next question is which part of it works. Cut three variants that each change one thing (opening line, opening frame, the claim), and run them against A in the same ad set so Meta arbitrates.
45–54 costs double, Instagram costs 8% more — both true, neither worth acting on. Meta has already moved 88% of spend to 55+ on its own, and narrowing would shrink the pool to save about $12. The age finding belongs in the copy brief, not the ad set.
Eight taps at $13.36 is the number that eventually decides whether this is a business, and right now it is unreadable — the sample is too small, and the door is painted, so a real card field will cut it substantially. It needs roughly 50 taps, about $650 more at current rates, before it means anything.
What the numbers are and where they can mislead.
1761459111652871, lifetime to date. Pulled 2026-08-09 ~14:50 PT.