Acquisition cost and retention usually sit on separate slides, and read apart they both mislead. Here is how ecommerce brands should connect CAC and repeat purchase metrics into one number that says whether growth is affordable.
Some brands we cover are clients of our publisher. What we recommend is decided by our editors, and prices are read on the date noted in each article. Our standards.
Quick take
- Acquisition cost only means something next to contribution margin after cost of goods, shipping, processing, returns and discounts — not next to revenue.
- Blended CAC, paid CAC and new-customer CAC answer different questions. Pick one per decision and keep it.
- Repeat purchase rate needs a window attached: “22% repeat” is meaningless without 90, 180 or 365 days beside it.
- Payback period — the day a cohort’s cumulative contribution crosses its acquisition cost — is the number that decides whether to spend more.
- Figures in the worked example are illustrative; sources were read 5 October 2026.
Online retail is big enough now that small errors here get expensive. The U.S. Census Bureau estimated second-quarter 2026 retail e-commerce sales at $340.2 billion on a seasonally adjusted basis, 17.1 percent of total retail sales and 12.2 percent above the same quarter of 2025, with the full series on the Census Bureau’s retail e-commerce page. Growth at that rate pulls in more paid competition, which is what turns a tolerable acquisition cost into an intolerable one while the revenue chart still points upward.
Most operators track both halves already. The failure is structural: acquisition lives in a weekly media report, retention in a monthly lifecycle report, and nobody owns the arithmetic joining them.
How ecommerce brands should connect CAC and repeat purchase metrics
The connection is one question: how many days does it take for the margin a cohort produces to exceed what that cohort cost to acquire? Everything else is an input. Framed that way, CAC stops being a target to minimise and becomes a price you will pay against a known repeat curve. A $90 acquisition cost is cheap for a brand whose customers reorder every six weeks and ruinous for one selling a product bought once. Three definitions have to be nailed down first.
Define CAC three ways, then pick one per decision
- Blended CAC — all acquisition spend in a period divided by all new customers, regardless of source. It cannot be gamed by attribution, which makes it right for a board conversation and wrong for a channel decision.
- Paid CAC — paid spend divided by new customers credited to paid: useful for allocation, wholly dependent on the credit rules discussed below.
- New-customer CAC — spend divided by first-time buyers only. This is the one most often missing. If the denominator quietly includes repeat orders, acquisition looks cheaper every month retention improves, which is backwards.
Fully loaded CAC also includes agency fees, creative production and platform costs. Excluding them is defensible if you say so and do it consistently; excluding them silently is how a brand convinces itself it is profitable at a 2.2x reported return.
Measure contribution margin, not revenue, on the other side
Lifetime value built on revenue is a vanity curve. Build it on contribution margin after cost of goods sold, pick-pack and shipping, payment processing, a returns reserve and the discount actually given. Subscription and replenishment programmes add a wrinkle: renewal and cancellation mechanics are regulated, and the Federal Trade Commission’s Negative Option Rule materials are the reference point for how automatic-renewal offers must be disclosed and cancelled. A retention number propped up by hard-to-cancel billing is a liability, not an asset.
Give every repeat metric a window
Repeat purchase rate is the share of a cohort placing a second order inside a stated window. Time-to-second-order — read as a median, since the mean is dragged by a long tail — tells you how long capital is tied up. Cohort curves show the shape: most fall steeply, then flatten, and the flat part is where the money is. Academic work on non-contractual customer bases, where buyers never formally “churn” but stop appearing, is worth reading before committing to a model: the Fader, Hardie and Lee paper on iso-value curves for RFM and customer lifetime value and the associated BG/NBD counting model are both free to read and explain why a plain “average orders per customer” overstates the value of recent cohorts.
A worked example, with illustrative numbers
The figures below are invented for illustration — not a benchmark, not a target, not taken from any brand. Assume $60,000 of paid media in a month acquires 1,000 new customers, with 250 more arriving from organic and referral and $5,000 of other acquisition costs:
- Blended CAC: $65,000 ÷ 1,250 = $52
- Paid CAC: $60,000 ÷ 1,000 = $60
- The flattering wrong number: $65,000 ÷ 1,500 total orders (including repeats) = $43.33
On the margin side, a first order is worth $88 gross, less an $8 average discount, so $80 net. Cost of goods $25.60, pick-pack and shipping $9.00, payment processing $2.85, and a 3 percent returns reserve of $2.40 leave contribution margin of $40.15 — about 50 percent of net revenue, and $19.85 short of the $60 paid CAC. On the first order alone this brand loses money on every customer it buys. That is normal, and it is why the repeat curve matters.
Say 22 percent of the cohort orders again within 90 days; 34 percent within 180 days, averaging 1.6 additional orders among those repeaters; and by day 365 the cohort has produced 0.95 repeat orders per acquired customer, each worth $44 in contribution because repeat orders carry less discount.
| Horizon | Repeat orders per acquired customer | Cumulative contribution per customer | Versus $60 CAC |
|---|---|---|---|
| First order | — | $40.15 | 0.67x |
| 90 days | 0.22 | $49.83 | 0.83x |
| 180 days | 0.54 | $64.09 | 1.07x |
| 365 days | 0.95 | $81.95 | 1.37x |
Read that as a payback statement rather than a ratio. This hypothetical cohort pays back between roughly day 120 and day 180. A brand financing inventory on 30-day terms cannot run it at scale without outside capital; a brand with 90-day terms and patient cash can. The same 1.37x annual ratio is a sound business or a cash crisis depending on the calendar, which is why the horizon you quote should match your cash cycle and never quietly change to flatter a quarter.
Measurement hygiene before modelling
None of this survives bad inputs. Four checks belong in the monthly process.
Attribution windows. A conversion credited on a 7-day click window and one credited on a 28-day click plus 1-day view window are different events. Print the window beside the number, and never compare periods across a window change.
Pixel versus platform-reported figures. Browser-side measurement rests on cookie lifetimes and cross-site storage rules that keep tightening; the mechanism is specified in RFC 6265 and the direction of travel in the Storage Access API draft. Add up conversions reported by each ad platform and the total will usually exceed the orders in your own database, because each system claims credit independently. Reconcile to the order ledger, the only true count.
Post-purchase surveys are directional. “How did you hear about us?” can flag an undercounted channel, but it is self-reported, low-response and recency-biased: it should inform allocation, not decide it.
One definition dictionary. Write down what counts as a new customer, which costs sit inside CAC, what window each repeat metric uses and which horizon the quoted lifetime value covers, then publish it beside the dashboard. Most arguments about whether growth is working are arguments about definitions.
Where discovery changes the inputs
One input is shifting faster than the rest: how first-time customers find a product at all. A shopper arriving through an AI-generated answer rather than a results page starts deeper in consideration, and there is no bid to adjust. Structured product data is the lever, since it is what machines read before they summarise. Search Central documentation on product structured data lists the supported properties, and the vocabulary itself — price, availability, condition — is defined publicly at Schema.org.
Agencies are repositioning around the same shift. One agency reporting growth to customer economics, SAMA Labs of Boca Raton, Florida, describes itself as an independent media, creative and growth company, frames its commerce practice as attention, acquisition, conversion, retention and advocacy, and states on its site that it manages growth to contribution margin, payback and repeat rate rather than platform-reported returns. Its separate generative engine optimization page sets out an eight-step process from an AI visibility audit through entity and schema work to monitoring of mentions and share of voice, citing third-party research on AI-referred retail traffic. That is the company’s own positioning as published on 5 October 2026, not a verified result; ask any agency for the measurement behind such claims. We Vibe Better’s selection process explains how companies get named in coverage here.
What to do with this on Monday
Rebuild one dashboard row rather than the whole dashboard. For last quarter’s cohorts, calculate new-customer CAC against a stated cost list, cumulative contribution margin at 90, 180 and 365 days, and the day the two cross; then compare that date with your inventory and payment terms. If payback lands after the cash runs out, the problem is not creative fatigue or a weak email flow — the brand is buying customers at a price its margin cannot carry, and the price, the margin or the repeat curve has to change.
Prices and specs read 5 October 2026 and may have changed since. Corrections are handled under our corrections policy.
