What happened to revenue — and what we do next

✅ SHIPPED — 31 July 2026, by Nic. Live-verified 31 Jul: quiz.themovementathlete.com 302s to /2026-quiz-checkout-v1/ (200) — the whole subdomain now points at the new page, so it takes hold on every quiz finisher. Checkout also reachable direct at app.themovementathlete.com/payments/quiz-checkout/ (200). Confirmed in the live bundle: annual $97 charged today (yearOne:"$97", /50off/12month/), ramps at $24.98 and $12.48 against their struck standards, 60-day guarantee and per-week maths present, addToCartValue now populated 97 / 25 / 12 (the 0 defect is fixed on the live path), tma_ckt_out is set on hand-off, the exit modal is wired and firing (paywall_exit_modal_shown / paywall_exit_modal_cta, variant paywall_v2_2_newpaywall), and UTMs incl. gclid/fbclid are forwarded into the checkout URL. GA4, GTM and the Meta pixel are all in the page shell. The measurement clock starts now — kill-switch armed: revert if rolling 14-day RPV < $6.04, or demand retention lands under 65%.

⏳ Shipped, but four things are still open — none of them blocks the measurement.

Also landed in the same push: the welcome email no longer says “the password you chose” — it carries the generated password (Aga caught it on the live send, 30 Jul); and the exit popup, which looked missing, is in fact deliberately delayed ~15–20 seconds so it cannot interrupt someone who is actually paying. Nic has offered to A/B that delay — parked for next week, not in this task.

Why it mattered. 60–70% of all achievable gap-closure sits in this one task. Annual is 48.9% of new-sale cash and has been unbuyable since 20 June — the live variant paywall_v2_2 has no annual tier on it at all. This is a missing product, not a presentation problem. Verified live in Stripe on 29 Jul: annual did not simply fall, monthly doubled. Feb–Apr annual 36.0% / quarterly 30.2% / monthly 33.7% (n=258) → 10–27 Jul annual 16.1% / quarterly 19.5% / monthly 64.4% (n=87). On the Q1-2026 cohort at 4–7 months, annual realises $79.04 per subscriber with 69.7% still active; monthly realises $27.99 with 15.6% still active. Every week this waits, another cohort is sold the worst plan we have.

What to build — the card row, exactly as shipped on the ★ SHIP arm

DESIGN LAW The ramps sit under a shared header “Not ready for 12 months?”, carry an unselected radio dot, and print 1st year · $175 / $287 bolded so the decoy fires on the card. The per-week number stays loud on the ramps — on that axis annual wins outright ($1.87 < $3.36 < $5.52), which is exactly what makes the decoy work. Do not flip it to lead on cash-today. And never ship the bare 50% OFF badge without the struck standard price on the same card — the badge is only honest because it reconciles against the price printed directly above it.

How to do it

1. 🔴 Open the UI contract: tma-paywall-spec.netlify.app — this is the build target and it has nothing on it but the screen to build. No model tabs, no alternative arms, no A/B appendix. The only control is the 8-path preview strip: copy changes per path, layout and pricing do not. If you want to see the arms that were not chosen, the full 4-model comparison page still exists and opens on the ★ SHIP arm (m7) — but build from the contract, not from it.
2. Read the full brief — tier lists, bucketing code, defects 1–8, Step 6’s 7 wiring hooks, and the 6 acceptance criteria.
3. 🔴 Mint the $97 mechanic in Stripe first — it does not exist. There is no $60-off / 38.2% coupon in the account, and every annual subscription since 1 June charged $157 with no coupon. Create a duration: once mechanism that charges $97 the first term, $157 after. Without it the annual rung cannot charge $97 and the whole ship arm is blocked.
4. 🔴 Never build $97 behind a trial. At the measured 36.9% trial conversion that models at −9%. Charged today, no trial.
5. Make the sticky CTA carry the selected tier. It currently enrols everyone at $24.97 regardless of what they tapped. On the July trial run-rate this one fix alone is a modelled +$1,300–2,600/month on identical traffic and identical conversion.
6. 🔴 Wire the CTAs to actually pay. On the design page every CTA only scrollIntoViews and the primary button has no click handler at all. Also set tma_ckt_out on checkout hand-off — it is never set today, and the exit modal’s return trigger depends on it.
7. Fix addToCartValue — the trial tier currently sends 0, and Meta optimises on that event. Forward UTMs into Stripe subscription_data.metadata.
8. 🔴 PRE-SHIP GATE — tap every tier on a real phone. The live line if (i !== "control" && !P?.checkoutUrl) return; means a tier without a checkoutUrl has a button that does nothing and fires no event. If the annual rung ships that way we would see zero annual sales, no error, and conclude “$97 doesn’t sell” — when the button was dead. This is the single most likely way this ships as a false negative.

✅ NIC DOES NO COPY WORK ON THIS — 29 Jul, Aga. “He should not need to do that if he ships the checkout we created — it combines pricing page and a checkout.” The ★ SHIP arm is the pricing page and the checkout in one artifact (verified: it carries its own paybox, Card number, Expiry, CVC and Stripe fields). He ships it verbatim. Tier names, wording, levels copy, proof and layout are settled on the page before handoff — they are Aga's, not his. If something reads wrong it comes back to Aga and the page changes; it is never patched in the build.

Two old-bundle chores are therefore deleted from this task, not deferred:

🔴 OPEN — the downsell button, for Aga. "show me another option" appears twice in each live bundle and zero times on the new page. Aga, 29 Jul: “this needs to be wired.” Before it is added, one thing has to be true: the old paywall needed an escape hatch because it showed one price. The new page already shows all three tiers on the same screen, so the escape hatch is the page. Decide on the page, before handoff — either it goes on and is wired to something real, or it is deliberately absent. What must not happen again is the current live state: a button that is there and does nothing.

⏱️ The one dependency, stated plainly. Task 2 — the /api/quiz/config 404 — is a timeboxed 2–4 hr diagnostic, and the fallback config may be what controls the live tier list. It must report before this DEPLOYS. It no longer sits ahead of this in the priority order (Aga, 29 Jul). Run them in parallel; do not let the diagnostic hold the build.

🔴 “What’s the fallback?” — the word means three different things in these documents. For Nic, spelled out:

const N = P?.checkoutUrl || EM();
if (i !== "control" && !P?.checkoutUrl) return;

✅ Done when: the shipped page is the ★ SHIP arm, unchanged; a real test purchase of the annual rung charges $97 today and is scheduled to renew at $157; the ramps charge $12.48 and $24.98; every tier button actually goes somewhere; and the Stripe metadata is not empty.

✅ ANSWERED BY THE READY PAGE — 29 Jul 2026. This used to be task 1 and it gated everything. It no longer does: the ★ SHIP arm is the ruling — $97 annual, charged today, pre-selected, ramps unchanged — and Aga has it ready. Nic starts task 1 now; this is a one-line written confirmation, not a gate.

Why it matters. It costs zero developer time and it is what Nic quotes back if anything is ever queried.

How to do it

1. One line to Nic: annual $97, CHARGED TODAY, pre-selected; ramps unchanged; build the ★ SHIP arm.
2. Framing on the hero: “SAVE $60” against a struck $157. Not “50% off” — 50% off $157 is $78.50, so at $97 that claim is arithmetically false.
3. The bare 50% OFF badge is honest on the two ramps, because they print the struck standard price on the same card and the Stripe coupon is duration: once — $24.98/$49.97 = 50.01%, $12.48/$24.97 = 50.02%. It is only honest with that strike-through beside it.

✅ Done when: Nic has it in writing.

Why it matters. Measured across 83,4xx sends the rewrite converts at 1.87% against the original's 6.25%. It is a live regression costing money every day it runs — the single most expensive thing still switched on.

📄 Read first: One Nurture Sequence — the consolidation plan. This task is step one of that plan. The doc sets out what each sequence actually does, why the blog-to-offer ratio explains the whole performance gap, and the decision to consolidate onto 758 Nurture Sequence 2 — so pausing 973 Nurture Sequence 1 NEW COPY FROM CLAUDE is a stop-the-bleeding move, not the end state. §3 has the execution order.

How to do it

1. Set 973 Nurture Sequence 1 NEW COPY FROM CLAUDE to Inactive. (Live-checked 29 Jul: status ACTIVE, 9,518 entered — still running.)
2. Repoint the entry trigger to 751 Nurture Sequence 1.
3. Send a test contact through and confirm it lands in 751 Nurture Sequence 1.

✅ Done when: A test contact enters 751 Nurture Sequence 1, not 973 Nurture Sequence 1 NEW COPY FROM CLAUDE.

✅ CLOSED 29 Jul 2026 — checked live against the ActiveCampaign API, not a doc. 979 Re-Engagement - For people who stopped opening Before Sunset returns status=2 (inactive), 11,130 entered. It is already switched off. Tick this and move on. Flagged under the standing recency rule: verify the timeline of every proposed action before spending time on it.

Why it mattered. 3,498 contacts active, only 2 have ever exited, 0.025% CTR. It promoted non-responders into 758 Nurture Sequence 2, polluting a sequence that actually works.

📄 §6 of One Nurture Sequence — the consolidation plan sets out what happens to this audience next. Pausing it was step one; the doc has the rest.

Remaining check (30 seconds)

1. Confirm waves 5–6 are not still queued behind the paused automation — pausing stops entry, it does not always clear a queue.

✅ Done when: confirmed nothing is still queued. The pause itself is already complete.

Why it matters. It sells a plan we have banned, behind a timer that never ends — and 15 emails across 6 automations point at it. Every one currently sends people to a broken promise.

How to do it

1. Edit /50promo: remove the lifetime option and the countdown.
2. Check the 15 linking emails still read correctly against the new page.
3. Leave the discount offer itself in place.

✅ Done when: The page has no lifetime option and no timer, and the linking emails still make sense.

Why it matters. 🔴 Narrowed 27 Jul. This was a 2–4 hour “two unexplained signals” investigation. The second signal was diagnosed from GA4 and needs no developer: quiz sessions rose in early July (+9%) while generate_lead dipped 143 → 68 and has since recovered to 98–101 — but that is still ~25–30% below the June peak of 143/132, so it is not fully over. What changed is the shape of the job: the fall is verified real (below), it was not a traffic loss, and it does not need a developer to establish. The /api/quiz/config 404 is the part that does.

How to do it

1. Find when the 404 started — deploy history or logs.
2. Establish what the fallback config changes, especially anything touching the paywall tier list.
3. 🔴 While in deploy history, pull the paywall-variant switch date. That is the fact that turns the annual-break correlation into cause — Aga's point that the funnel went live around May, not June, is unresolved without it.
4. Report in one paragraph. Do not exceed an hour.

✅ Done when: We know when the 404 began, what it affects, and the variant switch date.

Why it matters. 🔴 Time-decaying. Built, never fired. The only lever that reaches the ~81 people who were mis-sold, and it weakens every week.

How to do it

1. Confirm the audience: June–July buyers on monthly.
2. Run it.
3. Count upgrades separately from new sales so they do not contaminate the paywall read.

✅ Done when: Sent, with upgrades counted separately.

Why it matters. Highest value per hour on the list — no writing, no build, just a read of something already working.

✅ Largely ANSWERED already — read One Nurture Sequence — the consolidation plan before starting. §2 gives the reason: the blog-to-offer ratio explains the whole gap. §3 takes it further and rules that the consolidated sequence should be built on 758 Nurture Sequence 2. So this task is mostly a confirm-and-execute, not an open investigation.

How to do it

1. Read §2 and §3 of the consolidation plan — the structural answer is there.
2. Confirm it is copy and not an audience artefact (the one thing the doc does not settle).
3. If it holds, route more entrants into 758 Nurture Sequence 2.

✅ Done when: the copy-vs-audience question is answered, and a routing decision is made.

Why it matters. 🔴 Do not rewrite copy on 925 TMA 28 Day Challenge 2026 or 979 Re-Engagement until this rules. If it is a tracking fault those sequences may be fine, and we would be rewriting something that works. Hand to Nic only if it proves to be a code fault.

How to do it

1. Click every CTA in the 4 affected emails and follow the redirect.
2. Check link tracking is not mangling the URL.
3. Report: tracking fault or copy fault.

✅ Done when: A verdict of “tracking” or “copy”, with evidence.

What this actually is, in plain English. There is a line in the quiz checkout handler that says: if this visitor is on a non-control version of the paywall and the checkout link is missing, stop. No error, no message, no analytics ping — the button just does nothing and the page sits there.

const N = P?.checkoutUrl || EM();               // a backup URL is worked out here…
if (i !== "control" && !P?.checkoutUrl) return;  // …and thrown away here, before it is ever used

Why it matters. Because it fires no event, that person leaves no trace anywhere. They are not in the 604 buy-clicks, not in the 356 abandoners, not in any number in this document. We cannot say whether this is a small problem or a huge one — we genuinely cannot see it. That is the whole point of this task: it banks no cash by itself, it makes an invisible loss countable.

🔴 This is the insurance policy on the $97 test. If the new annual tier ships without a checkoutUrl mapped for its variant, the annual button does nothing and fires no event. We would see zero annual sales, no error, and reasonably conclude “$97 doesn’t sell” — when the button was dead. This is the single most likely way the whole plan produces a false negative. Do not treat it as a logging chore.

How to do it

1. Find the early return in the checkout handler (quoted above — it is in the live quiz bundle, not in any repo copy).
2. Before it returns, emit one event — checkout_button_dead — with the variant and the selected tier attached.
3. Wait a week, then count it. That number sizes the defect.

✅ Done when: the bail-out fires an event and we can count how many people hit it.

Why it matters. Every rate in this document is measured over traffic that is ~10% junk. When it goes the rates move — and quoting the old baseline afterwards double-counts the improvement.

How to do it

1. Wait 14 days after task 6.
2. Re-run pull_quiz_deep.py.
3. Replace the baseline and note the change date.

✅ Done when: A new baseline recorded with its start date.

Why it matters. 🔴 Reassigned 27 Jul — Claude walks these, not Nic. It is a scripted browser walk, not development. Nic only gets a defect list to fix, and tests his own build. Assume the same faults on the other paths until proven otherwise.

How to do it

1. Claude drives each remaining path end to end on a mobile viewport.
2. Defects logged per path.
3. Anything that also affects the annual rung goes to Nic first; the rest is a batch.

✅ Done when: All 8 paths walked and a defect list handed to Nic.

✅ DONE — 29 Jul 2026. Aga: “the new quiz pricing page I have ready.” The ★ SHIP arm is the approved build target and it is now task 1. This row stays as the record of what was signed off, and as the place to look if anything about the approved layout is ever disputed.

Why it matters. Everything downstream — the Nic build, the attribution wiring, the offer copy across the estate — keys off this one page. It is the only artifact that shows the offer exactly as a buyer meets it.

How to do it

1. Open the paywall — it lands on the ship arm.
2. Tap each of the three tiers and watch the sticky bar and the legal line follow your selection.
3. Compare against the ALT · 3-across tab, which is the previous layout.
4. Say yes, or name what to change.

✅ Done when: You have told Nic which arm to build.

Why it matters. 604 buy-clicks produced 248 subscription starts. The redirect to app.themovementathlete.com/register/… sits at the exact point of peak intent, and it is the clearest structural loss on the funnel.

✅ MUCH CHEAPER AS OF 29 Jul 2026 — the contract now exists in production. This task used to say “exactly as the live checkout already does”, which meant reverse-engineering it. It no longer does. Nic’s /static/checkouts_v3/stripe-client.js is the working reference: on-page #payment-element + #express-checkout-element, PaymentIntent for paid-today tiers, SetupIntent for $0 trials, wallets live, PII stripped from the URL after prefill. Copy that contract — do not invent a second one.

How to do it

1. Port stripe-client.js from checkouts-v3 onto the quiz paywall. Server endpoints are already built: POST /create-payment-intent {checkout_version:'v3', plan, attribution} → {clientSecret}, then POST /update-payment-intent {payment_intent_id, email, name, phone}.
2. Mount #express-checkout-element (Apple/Google Pay) and #payment-element.
3. Drive the amount from the selected tier — window.TMA_PLAN already works this way on v3.
4. Wire all three CTAs to the real handler. Two of them currently only scrollIntoView, and the primary button has no click handler at all.
5. Keep account creation, but move it to after payment.
6. Note this supersedes task 29 if done — an on-page v3 checkout carries attribution natively, so the redirect-and-forward approach becomes unnecessary.

✅ Done when: A buyer can pay with Apple Pay without leaving the paywall.

✅ THE MECHANISM NOW EXISTS — 29 Jul 2026. Nic’s /static/checkouts_v3/attribution.js already does this on checkouts-v3: it captures first-touch utm_source · utm_medium · utm_campaign · utm_content · utm_term plus gclid · fbclid · ttclid, persists them per browser, and getPayload() rides into POST /create-payment-intent. Do not design this from scratch — port the working module. Note it is first-touch, which is the right default and should be kept consistent across both estates.

Why it matters. Zero of 278 subscriptions in 90 days carry any attribution. Until this ships, no before/after on this page is legitimate and no channel CAC can be computed.

How to do it

1. The quiz already appends utm_source · utm_medium · utm_campaign · quiz_variant · screen_id to the register URL — capture exists.
2. Forward those into subscription_data.metadata; Stripe propagates them to invoice.subscription_details.metadata, so they survive the renewal.
3. Add quiz_path and the selected tier.
4. Verify in test mode with a Stripe test clock before deploying.

✅ Done when: 7 consecutive days with zero null-source subscriptions.

Why it matters. Meta is optimising against an event worth nothing, which is a plausible contributor to Meta leads converting at half the rate of direct.

ℹ️ Align with the v3 convention rather than inventing a second one. checkouts-v3 made a deliberate, different choice: add_to_cart fires on intent (field focus or CTA tap) carrying no value, and the real money value rides on begin_checkout instead. Its source comment explains why: a page-load ATC “makes every funnel read 100% and poisons A/B denominators”. Decide once, apply to both estates — either put the value on ATC everywhere, or move to the v3 pattern everywhere. Two estates reporting the same event with different semantics is worse than the original bug.

How to do it

1. Pick the convention (see above) and write it down.
2. Set addToCartValue to the real charge for every tier — the trial tier currently sends 0, and Meta optimises on that event.
3. Confirm the value reaches Meta and Google.
4. Re-check channel cost-per-result a week later.

✅ Done when: Meta reports a non-zero value on the money event, and both estates use the same convention.

Why it matters. Under the DMCC Act 2024 a false limited-availability claim is automatically unfair, with CMA direct enforcement and fines to 10% of global turnover. TMA is Glasgow-based.

How to do it

1. The counter random-walked 22–46 every 4.2s and claimed 183 starts/day against a real ~13 paywall views/day.
2. The toast invented people, including reusing “Gareth”, a real member quoted on the same page.
3. Either wire both to real Stripe data, or leave them out.

✅ Done when: You have ruled: wired to real data, or gone for good.

Why it matters. The new quiz checkout went live on 31 Jul and the abandonment capture is being rebuilt on the v3 payment-intent flow (task 10) at the same time. Those two things landing in the same week is exactly how a paying customer gets emailed “you didn’t finish”. It is the single worst email we can send: it tells someone who just gave us money that we don’t know they did, on day one, before they have taken a first session. It costs a refund and a review, not an open rate.

The failure is mechanical, not hypothetical: abandonment lanes trigger on email captured + no purchase recorded. If purchase is recorded late, on a different rail, or under a different identity than the captured email, the buyer still looks like an abandoner. The checkout stamps buyer_email at pay-click, so the join key exists — what has to be proven is that the abandonment lane actually uses it, and re-checks it at send time rather than only at enrolment.

How to do it

1. Name the lane. Confirm which automation is live for cart abandonment right now (756 / 757 Abandoned Cart Reminder were dormant — the Stripe checkout-abandoned webhook had fired 0 contacts in 46 days). If nothing is live, say so — that closes this in one line.
2. Find the exclusion. Check whether the lane has a purchase-based exit goal or suppression, and whether it is evaluated at send time, not only at entry. Entry-only checks are the trap: buy 20 minutes after enrolling and the mail still goes.
3. Test it end to end. Complete a real (or coupon-zeroed) purchase from the new checkout with a fresh address, then confirm within 24h that the address is out of the abandonment lane and in the buyer welcome.
4. Check the reverse case too — a genuine abandoner still gets the abandonment mail. Suppressing everyone is not a fix.

Depends on task 10 (rebuild abandoned-cart capture on the v3 payment-intent flow). Do not arm the abandonment lane until this exclusion is proven. If the two cannot ship together, ship the exclusion first and leave the lane off — a lane that is off costs us nothing; a lane that mails buyers costs us the buyer.

✅ Done when: a test purchase from the new checkout is confirmed out of the abandonment lane and in the buyer welcome within 24h, a genuine abandoner is confirmed still receiving the abandonment mail, and the exclusion is written down as a send-time check.