Two different problems ran back to back. Through July, acquisition cost held flat near $97 while first-order basket value fell 20% — an AOV problem. In August the reverse: basket value steadied and cost per new customer broke to $150, because conversion, not media, gave way. August closes at 0.61 against a 0.80 plan. Five inputs close the gap.
| aMER = | New customers × First-order AOV |
| Meta + Google Ads + other |
The denominator is the whole argument. The same month reads three different ways depending on what counts as acquisition spend — which is why any aMER figure has to name its denominator before it can be compared to another:
| Denominator | August spend | aMER | Read |
|---|---|---|---|
| Meta only | $118,082 | 0.67 | Flatters the number — ignores every other acquisition dollar |
| Meta + Google Ads | $126,175 | 0.63 | The two platforms we can see and steer directly |
| Meta + Google + residual | $129,064 | 0.61 | CFO Model definition — the one the plan is written against |
This page uses the third throughout. The denominator was never documented in the model — it was reconstructed from Meta and Google Ads actuals and verified to three decimals against the model's own aMER row for every month January to July. Google Ads is 6.4% of it and has never been audited; the residual is affiliate and tooling spend, estimated, and is the one line still unverified.
| Input | August | Elasticity on aMER | What it means | |
|---|---|---|---|---|
| Numerator | First-order AOV | ~$91.74 | 1.00 | The only input with undiluted leverage |
| ↳ | Gross AOV | $109.80 | 1.22 | Amplified — discounts and returns don't scale with it |
| ↳ | Discount rate | 14.9% of gross | 0.18 | Per 1% relative cut. Was 8.3% in March |
| ↳ | Return rate | 3.3% of gross | 0.04 | Per 1% relative cut. Not worth chasing |
| Numerator | New customers | 863 | 1.00 | Two independent sources, below |
| ↳ | Meta-driven | 615 (71%) | — | Sets the 0.71 ceiling on every media lever |
| ↳↳ | Link CVR | 3.260% | 0.71 | Was 4.25% in July, 5.03% in January |
| ↳↳ | Link CTR | 1.062% | 0.71 | Held 1.06–1.27% all year. Not the failing layer |
| ↳↳ | CPM | $65.85 | −0.71 | Falls as spend rises in this account |
| ↳ | Non-Meta baseline | 248 (29%) | 0.29 | SEO, email, affiliate, direct — arrives regardless |
| Denominator | DTC spend | $129,064 | — | |
| ↳ | Meta | $118,082 | — | 91.5% of spend |
| ↳ | Google Ads + other | $10,981 | +9.3% | The gain from removing it entirely, if it isn't earning |
Read the elasticity column as: improve this input by 1%, and aMER moves by this much. Meta drives 71% of new customers, so conversion, click-through and CPM are each capped at 0.71 — while basket value moves aMER one-for-one. That asymmetry is the reason a cost-per-purchase ceiling alone is an incomplete instruction: CPP cannot see AOV. Elasticities are local, accurate for moves of 10–20%, and they compound rather than add.
| Month | Meta spend | New revenue | Meta aMER | 1st-order AOV | Meta CAC |
|---|---|---|---|---|---|
| June | $116,792 | $106,502 | 0.912 | $97.08 | $106.46 |
| July | $102,957 | $108,238 | 1.051 | $89.82 | $85.44 |
| August | $118,082 | $79,168 | 0.670 | $91.74 | $136.83 |
15% more spend, 27% less new revenue. And it is entirely cost per customer, not basket value — first-order AOV rose $89.82 → $91.74 while Meta CAC went $85 → $137, +60%. That +60% has two independent halves: Meta cost per purchase +33% (about 60% of the rise), and the non-Meta new-customer base halving from 492 to 248 (about 40%). The second half has nothing to do with the ad account — it only shows up as a worse-looking Meta CAC.
| Input | Jul → Aug | Cost impact |
|---|---|---|
| CPM | $74.21 → $65.85 | −11% |
| Link CTR | 1.220% → 1.062% | +15% |
| Link CVR | 4.254% → 3.260% | +30% |
Impressions got 11% cheaper and we bought 29% more of them. Media was a tailwind all month. Conversion gave it back and more.
| Source | New cust. | Sessions | Conv. |
|---|---|---|---|
| Social | −26% | +12% | −34% |
| Direct / none | −34% | +4% | −37% |
| Search | −17% | −33% | +25% |
Social delivered 12% more sessions and 26% fewer new customers. Traffic was never the constraint. Search is the opposite and separate problem — a third of its traffic gone while converting 25% better.
Social net AOV fell $103.07 → $93.51, −9.3%. Blended first-order AOV looked flat at ~$92, so this appears nowhere else on this page — but social orders are 99% new customers, so this is first-order AOV on the acquisition channel, and it dropped nearly ten points. Search held at $106 and direct rose. Only the channel that buys new customers declined.
| Input | From → To | aMER | Cumulative | |
|---|---|---|---|---|
| August 1–30 actual | 0.612 | |||
| 1 | Link CVR | 3.26% → 4.25% | +0.132 | 0.745 |
| 2 | First-order AOV | $92 → $102 | +0.083 | 0.828 |
| 3 | CPM | $65.85 → $61.00 | +0.050 | 0.879 |
| 4 | Link CTR | 1.06% → 1.15% | +0.056 | 0.935 |
| 5 | Non-Meta new customers | 248 → 310 / mo | +0.049 | 0.984 |
The dashed line marks the October plan. Four of the five moves are needed just to reach it — the plan is not cleared by the two easy ones. All five together land at 0.984, marginally short of the stretch. A sixth lever sits off this bridge: Google Ads has held at $7–13K a month all year while Meta halved, and is now 6–9% of the spend denominator. If it is not producing new customers below $92 each, removing it is worth up to +9% aMER on its own — larger than any step except conversion. It has never been audited.
Three definitions are in active use and they disagree by 37% on the same month. July 2026:
The dashboard version charges Amazon's ad spend against Shopify-only new revenue while excluding Google Ads altogether — two errors that partly cancel, which is why it has gone unnoticed. This page uses the CFO model definition throughout. On the dashboard's current math, July already cleared 1.00.
Ruled out: a Meta attribution artifact (Shopify's own rate fell independently, 3.11% → 2.61%); the 12 August campaign rebuild (shows nothing in the daily series); and a losing test variant (all three arms of the NAD+ Solo test sit at 1.75–1.96%, control included). Established: spend rose 18% per day while new customers fell 26% per day — a dollars-versus-count comparison that no session-measurement question can explain away.
Still open: how much of the rate decline is Intelligems test traffic inflating session denominators — the internal read is ~82%, which would mean the recoverable share of Step 1 is smaller than 4.25% implies. Four tests launched 18–19 August have not been isolated. And /pages/nad-solo converts at ~1.8% in a ~3% site while taking a quarter of sessions, which points at routing — a campaign configuration change, faster than any site rebuild.