11 Aug 2026 · every TMA figure pulled live the same day (ActiveCampaign · SendGrid · Stripe both rails · Meta Marketing API · GA4) · three specialist reviews plus a check pass that re-derived the numbers and looked for errors in them. Revised twice; §8 records what we got wrong on the way. Revised again 11 Aug PM: the strategic judgement now leads, §2 works through the three cited case studies against their own denominators, and §4/§5/§7 were re-derived — which resolved the ad-spend figure and corrected four of our own numbers (§8).
What this strategy is worth to a company of TMA's size, and what would have to change for it to turn on.
The strategy is right. The tool is a scale play, and TMA sits below the scale where it works — but the thing holding it below that line is conversion, not traffic and not the vendor.
That premise is sound, it is not seriously contestable, and TMA's own history is the evidence for it. Every large cash event this company has ever had came from the same move: the existing list, plus a dated priced offer. The July lifetime sale took $19,997 in 11 days (72 sales, 67 distinct buyers, 2–12 Jul at $297); Black Friday 2025 took ~$31.6K on the same instrument. Owning the relationship and asking it for money demonstrably works here — so the strategy behind this proposal is not in doubt at TMA.
What we do not have is a measured rate. Those were one-off dated events, and nobody can currently say what fraction of an owned list converts on a given ask — the figure has never been isolated. 993 SALES SEQUENCE, armed 10 Aug, is the first always-on automation in the estate that asks for the sale, and its first clean read (~10 Sep) is the first measured conversion number we will have. That matters here because the whole proposal turns on what an email is worth to us once we hold it: the proven lever so far is asking the list we already own, and the unproven one is enlarging that list with cold strangers.
| Annual revenue | Monthly visitors | Revenue per web visit | Price per identified email | |
|---|---|---|---|---|
| Cymbiotika | ~$150M | — | — | $0.15 |
| MUD\WTR | ~$6.1M–47M | ~425–462k | — | $0.15 |
| Legion Athletics | ~$11.5–15.8M | ~245k | $4.60 | $0.15 |
| TMA | ~$108K | 30,730 8,438 US | $0.29 | $0.15 |
Cymbiotika is roughly 1,400× our revenue. Legion is 105–145×, and turns a web visit into $4.60 where we turn one into $0.29. Retention.com charges every one of these businesses the same $0.15 per identified email. The cost of the tool is flat; what it returns scales with what a visitor is worth to you. That is the whole reason it is an obvious yes for them and a marginal call for us — and it is not a criticism of the idea, it is a statement about which business it is being applied to.
An identified email costs $0.15. At TMA's measured monetisation it returns $0.02–$0.08. MODELLED from §2 The unit is upside-down by 2–8× before a penny of setup. It is only invisible today because Retention.com's free tier absorbs the first 1,000/month — which happens to be almost exactly our whole volume.
🔴 So scaling traffic makes this worse, not better. At 5× our traffic the fee is ~$660/mo against $105–425 of revenue; at 10×, ~$1,470/mo against $210–850. The free tier is not a discount here, it is the only reason the arithmetic survives at all.
The one route where it clears comfortably is route A in §2 — the assumption that a cold identified visitor buys a fitness subscription as readily as a supplement customer buys protein. That route returns $0.31–$1.23 per email, well above the $0.15. So the entire proposal reduces to a single question: do we convert an identified stranger like an ecommerce brand, or like ourselves? Everything measurable says the latter — most of all the funnel-order fact in §2.4, that we capture email twenty screens before the paywall, so the high-intent anonymous pool this tool exists to harvest is one we already own.
Stated as conditions rather than opinions. Each one is a way the unit clears $0.15 per identified email — the whole thing turns on that single number, so anything that moves it moves the verdict.
| Scenario | What would have to be true | How we would know | Realistic? |
|---|---|---|---|
| A · The modelled number is simply wrong | Identified visitors convert 2–8× better than the transplant predicts. Our $0.02–$0.08 is modelled from Legion's ratio, not measured on our traffic — and no case study publishes the time period that would pin it. | A free-tier trial measures it directly. 1,000 identities/month costs $0 in fees; the real cost is setup, the legal opinion and founder attention. | Most likely of the five — and the only one testable now |
| B · Our revenue per visitor rises | From $0.29 toward ~$1.16 (4×) to clear the fee, or ~$2.32 (8×) to clear fee plus setup. Legion sits at $4.60. | The quiz→trial→paid funnel, plan mix, and the 993 read from ~10 Sep. | Yes — and it is the work we are doing anyway. This scenario arrives as a by-product of fixing conversion |
| C · A higher-value product line | An offer with ecommerce-like AOV — equipment, supplements, a $100+ one-off — so an identified stranger has something to buy at first contact rather than a subscription to be convinced into. | Would follow a deliberate product decision, not a marketing one. | Not the business today. The $27 ebook is the only non-subscription SKU |
| D · The funnel changes shape | Email captured after the paywall instead of at screen 32 — which would manufacture the anonymous high-intent pool this tool exists to harvest. | Would show up immediately as fewer leads. | 🔴 Possible but a bad trade — it swaps a proven owned asset for an unproven bought one. Not recommended |
| E · US traffic grows substantially | More volume at the same or better monetisation. | GA4. | 🔴 On its own this makes it worse — past 1,000/month the fee starts and grows faster than the return. Only works combined with B |
The honest summary of the five: A and B are the live ones, and they point the same way. B is the work already underway, and it turns this on as a side effect. A costs nothing in vendor fees and would settle the argument with a measurement instead of a model — the case for running it is not the $0 fee, it is that it would replace our weakest assumption with a fact. What stands against it is the setup, the legal gate and the founder hours (§6) — which is a judgement about sequencing, not about whether the idea is sound.
The lever is conversion, not identification. Raise revenue-per-identified-email above $0.15 (≈$0.30 in year one, carrying setup) and this tool turns on by itself, with no new argument required. Leave it at $0.02–$0.08 and no volume of traffic, and no vendor, rescues it.
That is a dated, testable threshold rather than a refusal — and the first evidence lands ~10 Sep, from 993 SALES SEQUENCE, at £0. It is also why the two things adopted from this proposal (first-party suppression with a holdout, and the abandonment lane already running at $730–1,700/mo) are the same bet made where the conversion is already proven.
Where this could be wrong, stated plainly. The $0.02–$0.08 figure is modelled, not measured — it comes from transplanting Legion's ratio, and no case study publishes the time period that would pin it. If a real trial showed TMA converting identified visitors 4× better than modelled, the unit clears. That is exactly what a free-tier trial would measure, and it is the strongest argument for running one anyway — the counter-argument is not the $0 fee but the setup, the legal gate and the founder attention, priced in §6.
Four points came back on the first published version of this page. Three are factual claims about what it contained, so they were checked by counting rather than argued about. All three are correct, and one of them was understated.
| The point | What the check found | Where it is now |
|---|---|---|
| "Retention.com was barely mentioned" | Correct — and stronger than stated. It appeared zero times. The page said "the vendor" throughout, while pricing Retention.com's own published $0.15 per identified email and its 1,000/month free tier. The analysis was run on Retention.com's economics without naming it once. | Named throughout; §2, §6.1 |
| "The 3 specialist reviews didn't account for the success of Legion Athletics, MUD\WTR and Cymbiotika" | Correct. Those three names appeared zero times — not on this page, not in the growth review's §PA.10, not in any of the three specialist reviews. The proposal's own evidence base was never worked through. | Read at source, denominators rebuilt, transplanted three ways — §2 |
| "It didn't seem to account for most of the proposal" | Correct on the evidence sections. The method, the volume logic, the holdout, the legal gate and the four-group architecture were all engaged. The competitive evidence — the case studies, and the specific vendor named as the candidate — were not. | §2 in full; §8 records it as our error |
| "It seems this is backlog till November?" | Correct as the page read. It led with November and buried the rest. There are in fact three separate dates, the first of which is ~4 weeks away — plus one step that is not date-gated at all and could start this week. | §9.1 |
What changed as a result: the case studies are now sized against their own traffic and revenue and transplanted onto TMA. Doing that work reproduced the original finding from a completely independent direction — which is a stronger result than the first version had, and it was only available by taking the evidence seriously. The conclusion did not move. The basis for it is now considerably better.
| Question | Answer | Status |
|---|---|---|
| How many identities would it produce? | ≈810–1,350/month MODELLED | Answerable |
| Could the proposed experiment be read? | No — ~275 per arm against 640+ needed for even a large effect; 19,600+ for a purchase effect | Structural |
| What would it return? | ≈1.2 trials per 90 days on generous assumptions ⇒ ≈$1,667/trial against a $20 target MODELLED | Order of magnitude |
| What does the sending route touch? | Marketing, transactional and 1:1 all send from one mailbox on one domain — moving to hello@ changes the system, not the reputation | Verified §3 |
| What do Legion, MUD\WTR and Cymbiotika imply for TMA? added 11 Aug PM — the first version of this page did not answer this | Their figures are accurate. Sized against their own traffic and revenue, they imply $19–$85/month for TMA like-for-like — the same place our independent chain landed | Now answered §2 |
Where that points: the underlying question — does owning a relationship we already paid for beat buying it back? — is a good question, and it is testable on our own opted-in data with a real holdout, at no media cost and no legal gate. The three case studies point at the same mechanism, and TMA's version of that lane already exists, and is roughly an order of magnitude bigger. Full recommendation and the three dates in §9.
Three specialist reviews ran independently on the same verified dataset — measurement, economics, deliverability — each required to tag every number as measured, modelled or not knowable, with its source. A fourth pass re-derived the load-bearing figures and hunted for errors in the first three. It found several, including two of ours (§8).
Parts of the proposal's method are the reason this could be analysed at all, and they're being kept:
| In the proposal | Why it mattered |
|---|---|
| RB2B ruled out as B2B identity resolution | Correct — company and job-title data doesn't map to a consumer fitness subscription. |
| Refused to compute 8,438 sessions × 97% | Sessions aren't people, and that distinction is load-bearing here. Our own work this week hit the same trap from the other side: bot traffic from Singapore and China inflates one site's session count by ~21%, distorting every per-session rate drawn from it. |
| Treated vendor-reported ROI as unverified | Same standard we now apply internally — we withdrew a £2.41 cost-per-purchase figure last week after finding it was a short-window artifact; the full-life number for the same campaigns was £67.48. |
| Required a true holdout | Around a fifth of lapsed subscribers reactivate with no intervention, industry-wide. Without a holdout, a recovery programme counts those people as its own. |
| Made legal a gate, and flagged CAN-SPAM as insufficient for a UK company | The review reached the same conclusion — detail in §6.3. |
| Wrote kill criteria before starting | This decision is now registered the same way, with the conditions that reverse it (§10). |
The proposal cites three consumer-brand results from Retention.com's published library — Legion Athletics, MUD\WTR and Cymbiotika — as evidence that the mechanism is commercially deployed rather than experimental. The first version of this page did not work them through, and did not name Retention.com anywhere. That is the gap this section closes. The proposal's own caution is kept: these are vendor-reported, so they are used here as a ratio source, not as proof of return.
| Brand | Stated result | Mechanism described | Not stated anywhere |
|---|---|---|---|
| Legion Athletics sports nutrition | $116K incremental revenue | Captures abandonment events missed by Klaviyo and Elevar; converts with personalised discounts | Every denominator. No time period, no number of identities, no traffic base, no cost. Two of the three state an ROI multiple; none states what it was a multiple of. Without a period and a volume, none of these figures can be transplanted directly — which is why §2.2 rebuilds the denominators from public company data instead. |
| MUD\WTR coffee alternative | $100K+ incremental revenue · 15x ROI | Listed under the same identity-resolution product line | |
| Cymbiotika supplements | 10x+ ROI on the case-study index; $400K+ incremental in partner listings | Described as recovering revenue "we would have otherwise lost" |
Read live 11 Aug 2026 from retention.com/case-studies (17 studies listed) and the Legion study. CHECKED — the figures the proposal quotes are accurate · One note for completeness: Legion is not on the case-study index page, though its own study URL exists; MUD\WTR and Cymbiotika are on the index. Of the 17, the pattern is consistent — ROI multiples and revenue totals, never a denominator.
What $116K means depends entirely on the size of the business that earned it. Those figures are public:
| Brand | Annual revenue | Monthly visitors | AOV | Result as % of revenue |
|---|---|---|---|---|
| Legion Athletics | ~$11.5–15.8M | ~245,000 | ~$119 | 0.7–1.0% |
| MUD\WTR | ~$6.1M–47M sources conflict widely | ~425–462,000 | — | 0.2–1.6% |
| Cymbiotika | ~$150M | — | — | ~0.27% |
| TMA | ~$108K $9,019/mo net cash pace | 30,730 8,438 US | $97–157 | — |
Third-party estimates (Kona Equity · Grips Intelligence · Latka · press coverage), so treat each as a band rather than a figure. The bands are wide; the conclusion below does not depend on where inside them the truth sits. Cymbiotika is roughly 1,400× TMA's revenue; Legion roughly 105–145×.
Legion is the only one of the three with both a revenue figure and a traffic figure, so it carries the transplant. All three routes are run below, and the case is deliberately loaded in the proposal's favour where a choice exists:
| Route | Method | Implied for TMA |
|---|---|---|
| A · Legion's traffic ratio, unadjusted | $116K ÷ 245,000 monthly visitors, applied to TMA's 8,438 US sessions. Assumes TMA converts an identified cold visitor exactly as well as a $119-AOV supplement brand does. | $332–$1,330/mo low = $116K annual high = $116K over 3 months |
| B · Same, adjusted for how TMA monetises a session | Route A ÷ 15.7×, the net-cash monetisation gap (below). On a gross-revenue basis the gap narrows to ~13× and this row rises about a fifth, to ~$26–$102. | $21–$85/mo |
| C · Result as a share of the brand's own revenue | The 0.2–1.0% band applied to TMA's ~$108K/yr. | $19–$77/mo |
The honest limit on this arithmetic. None of the three case studies states a time period, so whether $116K was earned over a year or a quarter is not knowable from what Retention.com publishes. That single unknown is what makes route A span 4× on its own. The conclusion holds across the whole span — but the width is real, and no number in this section should be quoted as a forecast. MODELLED throughout
The monetisation gap, measured. Legion turns a web visit into $4.60 ($1,128,880 online sales ÷ 245,370 visitors, Nov 2025). TMA turns a session into $0.29 ($9,019/mo net cash ÷ 30,730 sessions). That is a gap of 15.7× as computed. One caveat in the proposal's favour: Legion's figure is gross sales and TMA's is net cash, so on a like-for-like gross basis the gap narrows to ~13× — which is the sensitivity carried in route B above.
Retention.com charges both businesses the same $0.15 per identified email. The identification step costs the same; the thing being identified is worth 13–16× less at TMA. That single ratio is what separates route A from routes B and C.
Where that leaves the case studies. Sized against their own denominators, the three results imply $19–$85/month for TMA on the like-for-like routes, and $332–$1,330/month only under route A's assumption that a cold identified visitor buys a fitness subscription as readily as a tub of protein.
The review's independent chain in §6.2 — built with no reference to these case studies — landed at ≈1.2 trials per 90 days. Priced out: $63/month is the absolute ceiling (1.2 × the $157 annual ÷ 3, i.e. assuming every trial converts and all at the top plan); at a realistic trial-to-paid rate it is nearer $31/month. The two methods agree — $19–$85 from the case studies against $31–$63 from the chain. The case studies, worked through properly, corroborate the finding rather than overturn it. That is the answer to whether they were accounted for: they were not, and they should have been — and doing so does not move the conclusion.
The ratio explains how much. This explains why, and it is checkable rather than argued.
All three case studies describe the same job: recovering someone who showed purchase intent but never gave an email. Legion's is explicit — abandonment events, converted with personalised discounts. Retention.com's flagship product line is reactivation of dormant and unidentified contacts; another study on the same page reports "8,050 dormant contacts recovered." In ecommerce that pool is large and valuable, because a visitor can browse, add to cart and abandon without ever typing an email address.
TMA's funnel is ordered the other way round. Email is captured at screen 32. The paywall is screen 52. The purchase is screen 56.
Quiz funnel skeleton: EMAIL is captured at screen 32; paywall view at 52; payment at 56. The tag quiz-email-captured fires when the email is typed, before any answer is submitted. CHECKED
So the high-intent anonymous pool that Retention.com exists to recover largely does not exist at TMA — because we already capture it, twenty screens before the paywall. What the vendor would reach at TMA is the population before screen 32: people who arrived and did not start or finish the quiz. That is a real audience, but it is by definition lower-intent than an ecommerce cart abandoner, and it is the group route B and C price at $19–$85/month.
Read as evidence about a mechanism rather than a vendor, the three studies make a real point: identifying and personally re-contacting someone who showed intent and left is a lane that reliably produces incremental revenue. That is well-supported and worth acting on.
TMA's version of that lane already exists — the abandonment machine, live since 7 Aug, running on people who gave us an email and then didn't buy. Its measured prize is $730–1,700/month MODELLED, against the $19–$85/month the vendor route prices at — between ~9× and ~90× depending which end of each range you take, so the direction is safe even though the multiple is not precise. Same mechanism, on an audience we already own, at no vendor cost and behind no legal gate. On the case studies' own logic, that is where the effort belongs first.
The follow-up point: rather than loading resolved addresses into ActiveCampaign, send them personally from hello@themovementathlete.com, the way the retention lanes already mail people outside AC.
Our first finding was that loading third-party-sourced addresses into ActiveCampaign puts the account at risk, since AC's acceptable-use policy covers addresses not collected by the sender. If the addresses never enter AC, that exposure doesn't exist. The point stands on its own terms and removes that finding entirely.
We pulled the from-address off the messages wired into the live automations, and the SendGrid account configuration. All three surfaces resolve to the same mailbox:
Sources: AC API from-address on messages 13674 / 13661 / 13662 (the 993 SALES SEQUENCE arc) and 11460 / 13071 · SendGrid API: themovementathlete.com is the account's only authenticated domain and hello@themovementathlete.com its only verified sender · the 1:1 rail's own operating rules. All read live 11 Aug. CHECKED
So the trade doesn't move. The marketing estate, the transactional mail and the personal rail are not separate senders — they are the same mailbox. Cold volume sent from hello@ shares its reputation with password resets, account verification, the buyer's login email, payment-recovery mail and the live sales sequence. Those are the messages where non-delivery has an immediate, direct cost.
Net effect of the route: it removes the ActiveCampaign policy exposure and leaves the deliverability question exactly where it was.
Three things would still need checking before that route could be costed NOT YET CHECKED: Google Workspace's own bulk-sender terms and daily recipient limits (AC's policy was binding on that route; Workspace has its own, and it deserves the same read); send mechanics, since the hello@ lanes run draft-only by standing instruction, so ~1,000/month is either ~1,000 drafts to approve or a new automatic arm; and reply load, since hello@ is a real inbox a human reads.
Unchanged by the route either way: the legal position — data-protection law attaches to obtaining and using the data, not to which pipe carries it — and the volume, cost and measurability findings below, which are properties of the audience and the funnel.
Starting from the 8,438 US sessions in the proposal — which is a 28-day figure, so ×30/28 = 9,040 per 30-day month, the base the chart uses — and applying five haircuts:
MODELLED — Retention.com's true match rate against our traffic is not knowable without a trial, and would additionally be distorted by tag-coverage gaps: pages served by WordPress code snippets bypass the theme header and load no container at all, so a reported match rate would be computed against instrumented pages rather than against US sessions.
Split four ways over 30 days, ≈1,100 identities/month gives ~275 people per arm. Against what a two-proportion test needs:
A two-arm design over 90 days reaches ~1,215–2,025 per arm — enough for the large proxy effect, still not for purchases. The 7- and 14-day reads in the proposal would return numbers well before any threshold. MODELLED from the volume above
Two structural notes: the "suppress from Meta" arm can't do much at ~$5/day of remarketing, since Meta needs around 1,000 contacts for reliable audience delivery and the overlap with our resolved pool would be minimal. And with four arms, each additional arm divides an already-short sample.
| Cost | Estimate | Note |
|---|---|---|
| Vendor fee — Retention.com, at its published $0.15/identified email | $0–225/mo | Free tier covers ~1,000 identities, which is where our volume lands. The fee was never the problem — the fixed costs below are, because there are ~90× fewer identities here to spread them over than at the case-study brands |
| Setup and integration | $1,500–3,500 MODELLED | Script, middleware to route identities through existing suppression checks, holdout randomisation, the email arc, legal opinion |
| Engineering | 8–16h | From a pipeline already committed to the break-even window |
| Founder time | ~1h/week × 13 | Against ~10h/week total |
Working the funnel at deliberately generous assumptions — 3,000 free-tier contacts over 90 days, cold-email open and click rates at the optimistic end, our measured quiz-to-trial rate applied to clickers:
≈1.2 trials. Against ~$2,000 of setup, that is ≈$1,667 per trial, versus a $20 cost-per-trial target. MODELLED
Break-even on setup alone needs a 3.3% cold-contact-to-trial rate — roughly 4–33× published base rates for cold consumer email. Re-run assuming the chain is five times too pessimistic, the gap is still ~83×. It's an order-of-magnitude distance, which is why it's answerable from arithmetic rather than from a pilot.
One measured detail bears on why. In the buyer-by-buyer join we ran on 10 Aug, 13 of the 15 buyers then on record had paid within about 20 minutes of first giving their email — median 14 minutes. The funnel converts in-session. That doesn't prove the non-identifiers never convert; it does mean the target population is defined by having chosen not to identify at the moment of highest intent, which is a difference to overcome rather than simply a group to reach.
Unchanged by the sending route, since data-protection law attaches to obtaining and using the data:
The proposal's economic engine is that we pay repeatedly to reach the same visitor. The measured state of that loop:
| What's running | What it produced |
|---|---|
| Every campaign that has ever produced a sale | Paused. What's live is direct lead generation — the person gives an email immediately — plus one ~$5/day warm test approved this week. In total that is not small: a live pull on 11 Aug shows $645 across the last 7 days (Meta $315 + Google $331), a run-rate of ~$2,766/month. The point is the composition, not the size — none of it is web-visitor remarketing, which is the line this proposal would displace. |
| Cold Meta, 98 days, ~$3,555 | ≈$444 per claimed purchase against a ~$76 break-even. No Stripe sale carries a Meta click ID. |
| Google, brand terms (£17.87) | 3 of the new quiz sales — people searching our own name. |
| Google, everything non-brand (~£343) | 1,569 clicks, no traceable sale. |
So the repeat-purchase loop isn't currently running, and the comparison the proposal sets up doesn't have a live counterpart to measure against. The relevant comparison becomes its cost against the £0-media routes to the ~24,000 people already in our database who gave us an email and haven't been asked for anything.
| Our claim | What happened |
|---|---|
| "We have never demonstrated that we can convert an email we already own" published in the strategic judgement, 11 Aug PM · corrected by Aga within the hour | Flatly wrong, and contradicted by our own ledger. Every large cash event TMA has ever had came from the existing list plus a dated priced offer — the July lifetime sale took $19,997 in 11 days (72 sales, 67 distinct buyers) and Black Friday 2025 took ~$31.6K. The true statement is narrower and less dramatic: we have never isolated a conversion RATE, because those were one-off events rather than a measured always-on ask. Corrected in the judgement above. 🔑 The error is the same shape as the one this page already records — a confident claim asserted without opening the artifact that holds the answer. |
| The first version of this page did not engage the proposal's own evidence found by Nic, 11 Aug | Correct, and measurable rather than a matter of opinion. On the first published version: Retention.com appeared zero times — the page said "the vendor" throughout, while costing Retention.com's actual published $0.15/email and free tier. Legion Athletics, MUD\WTR and Cymbiotika appeared zero times — not on this page, not in §PA.10 of the growth review, not in any of the three specialist reviews. And the timing answer led with November without separating the three dates that actually apply. The analysis had been done on Retention.com's economics without ever naming it, and the case studies were never worked through at all. Fixed in §2 and §9.1. Working them through moved the conclusion by nothing — but that is a finding, not a defence, and it was only available after doing the work. |
| "The same account and domain as the 83k list" | Correct as first written — but we published a wrong correction of it in between. Checking the hello@ point, we censused the whole AC message table, read back domains like calisthenicsacademy.co and bodyweighttrainingarena.com, concluded the estates were separate, and wrote that up. Wrong instrument: the account holds 13,132 message records spanning years of other brands and non-campaign rows, so that table's domain distribution says nothing about what the live estate sends today. Reading the from-address off the messages actually wired into the live automations gives one answer — hello@themovementathlete.com — matching SendGrid and the 1:1 rail. Only the third pass used an instrument that could answer the question. |
| "25–50% chance of a reputation event" | Downgraded from modelled to plausible. Gmail's ~0.3% complaint threshold is published and the 0.2–1.0% cold-cohort range is industry-sourced, but that specific probability band was a judgement presented as a calculation. |
| Our live ad-spend figure | RESOLVED 11 Aug — the ~$897/month figure was wrong. A live pull (spend_daily_pull.py, both channels reporting LIVE) gives $645 over 7 days = ~$2,766/month, so the ~$2,310 figure was the closer one and the lower figure should not be re-quoted. §7 is unaffected in direction — it rests on which campaigns are paused, not the total — but §7 now states the real total rather than leaving the ~$5/day test as the impression. |
| The ~1.2-trials chain | ✅ Reproduced independently. |
| The power calculation | ✅ Confirmed, and found to have understated the shortfall — the original quoted the sample needed for a proxy metric rather than for purchases. |
The volume estimate, the trial chain and the reputation range are all modelled rather than measured, and shouldn't be quoted as facts. The findings hold because the distances are large, not because those figures are precise.
The question is fair, because the first version of this page led with November and buried the rest. There are three dates, and the first one is next month:
| Date | What lands | Why it is the gate |
|---|---|---|
| ~10 Sep 4 weeks away | The first clean read of 993 SALES SEQUENCE — armed 10 Aug, the first always-on automation in the estate that asks an opted-in contact for the sale. (The July and Black Friday lifetime sales were one-off campaigns to the same list, and both converted — see the judgement above.) | This is the number that most directly informs the proposal's core hypothesis. It measures what an opted-in email contact converts at. A cold, third-party-resolved contact converts at some fraction of that. Until we know the warm number, the cold number cannot be forecast in either direction. |
| November | The four-group suppression architecture with a holdout, on first-party emails, into the Black Friday build. | Not a queue position — a dependency. Suppression can only be measured against remarketing spend worth suppressing, and live spend is one ~$5/day warm test. There is currently nothing to suppress. |
| 15 Dec | The registered review of the vendor decision (DEC-2026-08-11), with two dated reopen triggers (§10). | Follows the November read, because the November test is the £0 version of the same question. |
And one thing that is not date-gated. If the judgement is that the vendor question should be settled rather than parked, the step that unblocks it is the legal opinion in §6.3 — UK GDPR by establishment, Article 14 notification, the DPIA, and the US pixel-litigation exposure. It is the one gate the proposal itself set as hard, it does not depend on any of the three dates above, and it can start this week. It costs counsel time rather than founder time, and it is Aga's call, not a fleet one.
| # | Do this | Why now | Cost |
|---|---|---|---|
| 1 | Build the four-group audience architecture for paid suppression — on first-party captured emails (quiz leads, magnet leads, customers), into the November Black Friday work | It's the proposal's own mechanism, delivered on data we hold with real consent. November is when there'll finally be remarketing spend worth suppressing. | Engineering time; £0 media, no legal gate |
| 2 | Run it with a real holdout — a randomly assigned slice that gets no suppression treatment — and read incremental purchases against spend avoided through our own Stripe attribution | This is the honest test of the whole idea. If owning the relationship beats re-buying it, the mechanism is proven on data we control. | Included in #1 |
| 3 | Put a monitoring heartbeat on every data sync | Priced by a finding from the same day: the Facebook lead-form → ActiveCampaign sync had been dead since 1 Dec 2025. ~2,100 paid-for leads expired unretrieved; ~264 are still inside the retrieval window. Any integration has that failure mode — it works on install day and stops quietly later. | Small, one-off |
| 4 | Hold the vendor question against its triggers (§10) rather than re-arguing it | Both triggers are dated. Nothing is lost by waiting for them, and the December window is the strongest fitness-intent month to read a cold arc anyway. | £0 |
If the vendor question does reopen, the only shape worth building is a fully isolated one: a separate sending platform on a separate domain — not hello@, which §3 shows is the shared mailbox — with cold-list hygiene, and contacts graduating into the main estate only by an explicit act such as completing the quiz or buying.
Registered with a review date of 15 December 2026. Two conditions would reopen the vendor question, and both are dated rather than debatable:
Separately, the §7 premise expires on its own terms if warm remarketing spend returns to scale and 993 SALES SEQUENCE — the first automation here that asks for a sale, armed 10 Aug, first clean read ~10 Sep — shows the email estate converts opted-in contacts.
| Still open | Who / when |
|---|---|
The hello@ route's own costs — Workspace terms and limits, draft-versus-send mechanics, reply load (§3) | Not yet checked either way |
| The ad-spend reconciliation, ~$897 vs ~$2,310/month (§8) | Assigned |
| Retention.com's real match rate against our traffic | Not knowable without a trial, and distorted by the coverage gaps in §4 |