FiftyStrong · paid social

Ad copy and the event to optimise on

Two creatives already cut, one destination: the diagnostic quiz at fiftystrong.app/quiz. Below is paste-ready copy for both, the button, and the conversion event to buy against — plus three things that have to be fixed before any of it spends.

Optimise for

Lead

Fires the moment she reaches the paywall, having answered ~25 questions. The only event deep enough to mean something and frequent enough to exit learning phase.

Button

Learn More

Highest click-through of the eligible set. With a deep optimisation event you want volume at the top for the algorithm to sort. Challenger: Get Started.

Pixel created, connected and firing. FiftyStrong Pixel · 1055150507004620, created in the Reelly portfolio and connected to ad account 1761459111652871 only. Deployed to fiftystrong.app/quiz and verified by walking the funnel end to end on production: all five events fired in order and the dataset's last_fired_time moved from never to the moment of the walk. (Event counts in Events Manager lag by up to 30 minutes — last_fired_time is the immediate proof of receipt.)

It is a separate dataset, not the existing Reelly Pixel. One dataset serving two products merges their events into a single optimisation model and a shared set of audiences; Reelly Pixel already carries live spend and a 7.1/10 quality score, so feeding it a menopause quiz would have degraded Reelly's targeting and given FiftyStrong a signal contaminated with video-editor traffic.

Moment	Standard event	Custom event	Verified on prod
Lands on the quiz	`PageView`	—	fired
Starting — answers Q1	`ViewContent`	`QuizStart`	fired
Finishing — reaches the paywall	`Lead`	`QuizComplete`	fired
Paying — taps the buy button	`InitiateCheckout` value 18.00 USD	—	fired

Her archetype is deliberately not sent to Meta. An earlier version of this wiring attached start_here (mobility / balance / strength / energy) and goal_match to the completion event, so creative could be read per segment. That is an inference about a named person's physical condition — derived from questions about stairs, floors and continence — attached to an advertising identifier. Meta's Business Tools Terms prohibit sending sensitive user data, and a balance tag amounts to telling Meta this woman is unsteady on her feet. Those parameters have been removed. The segment breakdown belongs in our own analytics; the pixel carries only that she finished.

Unrelated bug found and fixed while verifying. Screen 5.1 (the "putting your plan together" loader) has no button — its only exit was a requestAnimationFrame loop, which browsers pause in a hidden tab. A woman who switches to a text message mid-loader could come back to a stalled screen, and she would never reach the paywall, which is the exact event we are optimising on. The loader now runs off the wall clock, so a background pause is absorbed rather than restarted, and a setTimeout guard advances the screen even if rAF never resumes. Confirmed by completing a full walk with document.hidden === true — the condition that previously stranded it.

Decision one

Why Lead, and not the others

Four events are wired. The question is which one Meta buys against, and the binding constraint is volume: an ad set needs roughly 50 conversions a week to leave the learning phase and deliver predictably. Anything rarer than that and the algorithm never gets a stable signal, no matter how meaningful the event is.

Event	What it means	Est. of clicks	Clicks/wk for 50	Verdict
`PageView`	Landed on the quiz	~100%	50	Too shallow
`ViewContent` `QuizStart`	Answered question one	60–75%	~70	Buys tappers
`Lead` `QuizComplete`	Finished every question, reached the paywall	20–30%	~200	Optimise here
`InitiateCheckout`	Tapped the buy button	2–5%	~1,650	Too sparse

At a realistic $1.00–1.50 CPC for women 48–65 in the US, Lead needs roughly $35–45/day to clear 50 conversions a week. InitiateCheckout would need around $300/day to do the same — an order of magnitude more than a validation test should cost, and you would sit in learning phase the entire time.

There is a second reason, and it matters more than the arithmetic. InitiateCheckout currently fires on a fake door — the paywall button answers "Checkout isn't open yet… nothing has been charged." Optimising toward women who tap a button that cannot complete would train delivery on an outcome that will mean something entirely different once payments are live. Lead means the same thing before and after launch, so the audience Meta learns now stays valid.

One change made to support this. QuizComplete is a custom event, and delivery optimisation on a custom event starts with no cross-platform training data behind it. It now fires alongside the standard Lead, same moment, same parameters. Optimise on Lead; keep reading QuizComplete for reporting, because it carries start_here (her archetype) and goal_match — which is how you find out whether the balance ad and the energy ad are pulling different women.

Decision two

Button, headline, description

Meta's button list is fixed. Download is out — there is nothing to download and the App Store has nothing on it. Sign Up mismatches: the quiz asks for no account. That leaves Learn More and Get Started.

Run Learn More as the primary. It consistently earns the highest click-through of the two, and because delivery is optimising on a deep event, extra top-of-funnel volume is useful rather than wasteful — Meta sorts it. Test Get Started as the challenger once each has 50 leads.

Headlines

shown under the video · keep ≤40 chars

H1Low friction, names the time cost

Find your starting point in 2 minutes

H2Names the personalisation

A plan built around how you move

H3Question — pairs with the defiance cut

Where should you actually start?

H4Qualifier, no personal assertion

Made for women over 50

Descriptions

often truncated — front-load

D1Removes the three usual objections

At home · No equipment · 12-minute sessions

D2Free applies to the check, not the plan

Free 2-minute movement check

Creative A · "Says who?" · 14.8s

Captions

The cut opens on a woman being told it's too late and answers back. The caption's job is to catch the identity the video just created and hand it somewhere to go. The dashed line marks roughly where mobile truncates to "See more" — everything before it has to earn the tap.

Primary text

3 variants · test against each other

A1Defiance — closest to the cut

Somewhere along the line women get handed a script: the fifties are the decade you start managing decline. Strength, balance and steady energy don't have an expiry date. They respond to training — the programme just has to match the body doing it. Answer a few questions about how you move right now: stairs, floors, balance, energy. Two minutes, and you'll see where to start. At home, no equipment.

A2The thing nobody explains

Nobody explains that perimenopause changes how the body answers back to exercise. The workouts that always worked stop working. The advice everyone repeats starts to feel wrong. And the usual conclusion — try harder — is the one thing that doesn't help. The answer isn't harder. It's different, and it's specific. Take the 2-minute movement check and see which of the four needs attention first: mobility, balance, strength or energy.

A3Short — best for Reels/Stories

"It's too late to start." It isn't. Balance is trainable at 55, at 65, at 75. So is getting off the floor without using your hands. Two-minute check, then a plan that begins where you actually are. At home, no equipment.

Creative B · workout programme · 70s

Captions

This one is a demonstration — eight named moves, no equipment, a chair. Demonstration creative wants specific copy: numbers it can verify against what it just watched. Seventy seconds is long for cold traffic, so also cut a 15-second version from the first two movements plus the end card.

Primary text

3 variants · test against each other

B1Specificity — matches the demo

Eight moves. Twelve minutes. A chair if you want one. That's a FiftyStrong session — made for women over 50, done at home, nothing to buy and nothing to carry. Which eight you get depends on what your body needs first. Answer a few questions about how you move today, and we'll show you where yours should start.

B2Function-led — strongest evidence

Most exercise programmes for women over 50 pick one of two mistakes: treating them as fragile, or as though they're 25. This one starts by asking how you actually move. Can you get off the floor without your hands? Do you reach for the stair rail? How long can you hold one leg? Then it builds the sessions around your answers. Two-minute check, twelve-minute sessions, at home, no equipment.

B3Objection-led — widest top of funnel

"I don't know where to start" is the most common thing women over 50 say about exercise. So we start by finding out. A few questions about stairs, floors, balance and energy — then a plan that begins where you are, not where a 25-year-old would. Twelve minutes a session. At home. No equipment.

Before this spends

Three things that will get the ads rejected

None of the copy above trips Meta's policies — it never asserts anything about the viewer, and it makes no claim above what the evidence supports. The creatives are a different matter, and the account is inside the shared Reelly portfolio, so strikes here are not free.

1. Both cuts are before-and-after body transformations. Meta's health and wellness rules prohibit before-and-after imagery and "unexpected or unlikely results" outright. Creative A opens on a midsection in the "before" state and resolves to a lean "after" — that is the prohibited pattern in its clearest form. Creative B ships INSET_BEFORE_AFTER, a fabricated belly pair, which its own README already flags. Both women are AI-generated and neither transformation happened, which is separately a problem under UK DMCC rules if you run outside the US.

Fix: flip INSET = INSET_NUMBERS in render_overlays.py and rebuild — the safe card (8 MOVES · 12 MINUTES · NO EQUIPMENT) already renders every run. For Creative A, cut the bathroom "before" and open on the line instead: the defiance lands on the words, not the body.

2. Two lines assert things about the viewer. Creative A's "you're 56 / it's too late to change" states the viewer's age — Meta's Personal Attributes policy prohibits copy implying you know a person's age or health status, and quotation marks are not a reliable defence at automated review. Creative B's 8 moves to blast menopause belly both names a health status and trades on negative self-perception.

Fix: "it's too late to change" — says who? keeps the whole idea with no age in it. For B: 8 moves · 12 minutes · no equipment, or Strength training made for women over 50.

3. Creative B's end card now contradicts the funnel. It reads "Get your customized fitness plan on us" over an App Store badge. The plan is $18/month, not on us; the destination is a quiz, not the App Store; and the App Store has nothing to download. An ad promising a download that lands on a questionnaire is both a conversion killer and a landing-page-consistency flag.

Fix: replace the badge with the quiz CTA and change the line to Find your starting point over 2-minute check · at home · no equipment.

Setup

Destination and order of operations

https://fiftystrong.app/quiz?utm_source=meta&utm_medium=paid&utm_campaign={{campaign.name}}&utm_content={{ad.name}}

Point ads at /quiz, never the root — the root is a coming-soon page with a dead App Store badge, which is the worst possible landing for paid traffic.
Never ship a link containing ?dev=1: that reveals the screen-stepper HUD.
~~Create the dataset, connect it to the ad account, deploy and verify.~~ Done — 1055150507004620, live on /quiz.
Fix the three creative problems below. They are now the only thing between you and spending.
Objective Leads → conversion location Website → conversion event Lead. Broad targeting, women, 48–65, US. Let the event do the targeting; interest stacks on this audience mostly narrow you into a worse CPM.
Start with the two creatives × three captions in one ad set, not six ad sets — you need the conversions concentrated to escape learning phase.
Revisit InitiateCheckout as the optimisation event once payments are live and you are clearing ~1,600 clicks a week. Not before.