01Three questions, three different answers
The campaign is being read as one result. It is three, and only one of them is still open.
Can you get her attention?
8.31% link CTR, $30 CPM, $0.36 a landing-page view on US women 48–65. That is cheap traffic in a demographic that is supposed to be expensive.
Does the message hold her?
74% of the women who touched question one finished all nineteen and reached the paywall. Frequency 1.05, CPM falling, cost per paywall improving every day.
Will she pay?
Zero dollars, because the checkout button was a painted door that returned a fake error. 12 women tapped it anyway. That is the entire evidence base for revenue.
So the honest headline is not the ads worked and not the ads failed. It is: everything upstream of money is proven, and money has never been tested. The next dollar should go at exactly that seam and nowhere else.
02Where 6,485 impressions went
Bar width is the share that survived the previous step — the funnel's actual information. Counts are absolute.
CPM
It was still getting better when it stopped
| Day | Spend | CPM | Paywall views | Cost each | Checkout taps |
|---|---|---|---|---|---|
| Fri 7 Aug | $41.17 | $37.74 | 27 | $1.52 | 2 |
| Sat 8 Aug | $40.53 | $39.77 | 49 | $0.83 | 4 |
| Sun 9 Aug | $68.85 | $26.75 | 78 | $0.88 | 6 |
| Mon 10 Aug | $44.75 | $24.85 | 55 | $0.81 | 0 |
CPM fell 34% across four days and cost per paywall view fell 47%. Frequency never exceeded 1.27. Nothing about this campaign was exhausted — it was paused with room left, which means the traffic side does not need to be re-solved before the next test.
03One ad did nearly all of the work
| Ad | Spend | Paywall views | Cost each | Checkout taps | Cost each | Watched to 95% |
|---|---|---|---|---|---|---|
| A · Says Who15 seconds | $136.67 | 161 | $0.85 | 11 | $12.42 | 28.6% |
| B · Workout Programme70 seconds | $58.63 | 48 | $1.22 | 1 | $58.63 | 7.4% |
A is 1.4× cheaper per paywall view and 4.7× cheaper per declared purchase intent. The gap widens the deeper you go into the funnel, which is the signature of a hook that qualifies rather than merely attracts.
The practical consequence: creative is not your bottleneck and should be frozen. Run A alone in the next test. Every new video you shoot before the money question is answered is spend on the one variable already working.
04The 209 women you can never contact
The quiz collects a first name for personalisation and nothing else. No email, no phone, no account. When the painted door threw its error, every one of those 209 women left and became unreachable.
That is not a rounding error. It is 209 qualified, self-declared members of the target segment — women who answered nineteen questions about stairs, floors, balance and energy — and the twelve who said they would pay. They are exactly the people the outstanding assignment from the August verdict says to interview, and exactly the audience a second offer would be tested against.
Fix this before the next dollar
An email field costs an hour and changes the return on every future test dollar. Without it each round of spend produces one number and evaporates. With it, each round also produces a list, a remarketing pool, and an interview pipeline.
Place it immediately before the price — where should we send your plan? — so it captures the full 209, not just the 12 who reach for a card.
05Does $18 a month survive its own acquisition cost?
This is the part worth being uncomfortable about, because it can be estimated now and it does not look good.
At $18/month you net roughly $15.30 through Apple's small-business rate or $17.18 through Stripe. Call it $16. Now run the cost of a customer against the observed funnel — $195.30 produced 209 paywall views, so every assumed conversion rate implies a specific CAC:
| If paywall → paid is… | Buyers per $195 | CAC | Months to repay CAC | Read |
|---|---|---|---|---|
| 1.0% | 2.1 | $93 | 5.8 | Dead at this price |
| 2.0% | 4.2 | $47 | 2.9 | Marginal — needs strong retention |
| 3.0% | 6.3 | $31 | 1.9 | Workable |
| 5.7% every painted-door tapper converts | 12.0 | $16 | 1.0 | Fantasy |
Painted-door taps do not survive contact with a card field. The usual range is that 25–50% of people who tap a fake checkout complete a real one. Apply that to 5.7% and the realistic landing zone is 1.4%–2.9% — a CAC somewhere between $34 and $70, against a subscription that returns $16 a month.
A monthly fitness subscription bought through a quiz funnel typically keeps 55–65% of buyers into month two and averages three to four months of life. That is $48–$64 of net lifetime revenue. Against a $34–$70 CAC, $18/month is somewhere between break-even and underwater, and you would wait four months to find out which.
The lever
Every competitor in this category — RH, BetterMe, Fortify — leads its quiz paywall with an annual price and uses monthly as the decoy. That is not a pricing preference. It converts a retention bet you must wait four months to settle into cash on day one.
At $89 for the year, a 2% conversion rate produces a $47 CAC against ~$86 net collected immediately — 1.8× on the first payment, before a single renewal. The same 2% at $18/month is a coin-flip that takes a quarter to resolve.
Which means the next test should not simply make the door real. It should make the door real and fix the price behind it, because a passing result at $18/month is still not a business.
06Test 2 — “does anyone actually pay?”
Budget $500. Roughly seven days, two to three of them build. Everything upstream of the paywall stays byte-identical so the 209-view baseline remains the control.
-
Wire a real checkout
Stripe Checkout with Apple Pay and Link enabled. Apple Pay matters disproportionately here — a 57-year-old will authorise with a thumbprint who would abandon a card form.
This is the whole point of the test. Everything else in this list is secondary to a charge landing.
-
Capture email before the price
One field, framed as delivery of her plan rather than as a signup. Ships with the checkout, costs an hour, and is the difference between a test that produces a number and a test that produces an asset.
-
Split the offer, not the creative
Cell A: the current $18/month, alone — an honest control against round one. Cell B: $89 for the year presented as $7.42/month, with $18/month sitting beside it as the decoy. Randomise 50/50 at the paywall.
Run Ad A · Says Who only. Same audience, same nineteen questions, same copy.
-
Spend $70/day for seven days
At round-one rates that is roughly 540 paywall views, about 270 per cell — enough to tell zero from five from fifteen, which is the decision at hand. It is not enough to rank the two prices precisely, and should not be read as if it were.
-
Read cash, not signals
Paywall → paid rate, CAC, average order value, refund requests inside seven days. The lead and checkout-intent metrics from round one stop mattering the moment a real charge exists.
Why not just build the app
A web checkout answers the money question in two days for $500. Shipping the iOS app answers the same question in three months, and only after App Review, signing, and a store listing — on a persona nobody has spoken to. The app is the right build; it is the wrong next step.
07Thresholds, written down before the test runs
Set now, on paper, so the result cannot be re-interpreted into whatever is convenient afterwards.
One caveat to hold onto: even a pass only proves acquisition. Month-two retention is the actual profit question, and no amount of ad spend can answer it. Leading with the annual price is how you stop needing to.
08Four things not to do next
- Do not shoot more creative. Ad A is working, unfatigued at frequency 1.05, and getting cheaper daily. New video optimises the only variable that is already solved.
- Do not scale the current campaign. More $0.93 leads you cannot bill and cannot email produce nothing but a larger bill.
- Do not broaden geography or audience. One proven cell is exactly what you want when the variable under test is price. Widening now adds noise to a small sample.
- Do not fix the 48% who never touch question one. It is the biggest remaining leak and genuinely worth money — later. Changing it during this test moves the baseline you are measuring against.
09Owed before you take a real card
The moment checkout is live you are selling a product that does not exist yet. That is normal pre-sale practice and it is fine — provided these are settled first, not afterwards.
- What arrives on day one. A buyer on Tuesday needs something real in her inbox on Tuesday. Decide the first deliverable before the button goes live.
- A stated delivery date and a refund guarantee. Written on the paywall, in plain words, and honoured without argument. At this volume a refund costs $89 and buys a conversation worth more.
- What $89 actually promises. The August verdict flagged this and it is still open: does the price commit you to a year of programming, or to fourteen days plus a published roadmap? Decide before you sell it, because five paying strangers is a real obligation.
- A health-claim pass over the paywall copy. Menopause-adjacent claims to a 50+ cohort, now attached to a charge rather than a quiz, is a different risk posture. The existing voice rules avoid the worst of it; the checkout page has not been reviewed.
- Name clearance. Still outstanding from the README — a USPTO knockout search on
FiftyStrongandFifty Strong, before money changes hands under the name.