The three CACs every operator confuses

A founder tells you their CAC is A$74. A board member hears A$74 and assumes it is the cost to win the next customer. The finance lead pulls the all-in number and gets A$210. The growth lead, asked whether to add another A$50k of spend, has no number at all. Three people, three implicit definitions, one word. The confusion is not pedantic. It decides whether you scale, hold, or pull back.

Customer acquisition cost is the most quoted unit-economics metric in ecommerce, and the most quietly ambiguous. There are at least three distinct quantities hiding behind the acronym. Each is correct for a specific decision and misleading for the others. Here is how to separate them, with the arithmetic, and why the third one is the number that should govern your next budget move.

Three numbers, three questions

Start with definitions, because the whole problem is definitional. The standard treatment from 1-800-D2C and the SaaS Metrics Standards Board separates blended from paid (new-customer) CAC. The third, marginal CAC, comes from the marketing-science literature on response curves and is the one operators skip.

The gap between blended and new-customer CAC is well documented and large. Take Shopify's own by-industry benchmark, which is at least primary-published rather than aggregator-relayed: it puts annual average CAC at US$377 for electronics, US$129 for fashion and for home and garden, and US$127 for health and beauty (US and global data, 2021, brands with fewer than four employees; Shopify, 2021). Yet the blended DTC figure most operators quote sits at roughly US$68 to US$100, with broad-category aggregators landing near US$78 (US data; Upcounting 2025). There is no clean Australian equivalent published to a primary source: the AU agency tables that circulate mostly re-label US WordStream data, and an Australian consultancy notes only that local CAC tends to run roughly 20 to 35% higher than US averages (Uncommon Insights), so treat the figures above as the shape of the gap, not local levels. Those numbers are not in conflict. They measure different things. The by-industry figures are closer to all-in retail cost per acquired customer; the US$78 averages in every channel that costs little or nothing. Quote one in a context that calls for the other and you will either talk yourself out of a healthy business or into an unhealthy one.

Why the marginal number is always the highest

Blended and new-customer CAC are both averages. Average CAC is comforting because it spreads the expensive customers across the cheap ones. The trouble is that growth decisions are not made on the average customer. They are made on the next one, and on a paid channel the next customer is reliably more expensive than the last.

This is not a quirk of any one account. It is the concave shape of the advertising response curve, one of the most replicated findings in marketing science. The meta-analysis by Sethuraman, Tellis and Briesch in the Journal of Marketing Research (2011) pooled 751 short-term advertising elasticities from 56 studies and found a mean short-run elasticity of 0.12 (Sethuraman, Tellis & Briesch, JMR 2011). An elasticity of 0.12 means a 1% increase in advertising lifts sales by roughly 0.12%. Response is sharply diminishing. Each extra dollar of spend reaches a less responsive slice of the audience, so it buys fewer incremental customers, so the cost of each incremental customer rises. That is marginal CAC climbing, and average CAC hides it.

The maths, worked

Marginal CAC is not exotic to compute. It is a first difference: the change in spend divided by the change in customers it produced, between two spend levels.

Now connect it to profit, because a CAC means nothing without the margin that has to clear it. The breakeven rule from contribution-margin economics is simple: a new customer pays for itself on the first order when the first-order contribution margin (revenue after landed COGS, fulfilment and payment fees) exceeds CAC. With an A$90 average order value and a 40% first-order contribution margin, each new customer throws off about A$36 of first-order margin. At the A$67 average CAC that looks underwater on order one and is rescued by repeat purchases. At the A$132 marginal CAC, step C only works if those customers come back: Shopify merchants see a repeat-purchase rate near 27% (US data; Shopify 2025), and the second-order economics carry the cohort. The decision is not "is our CAC good." It is "does the next customer's lifetime margin clear the marginal cost of winning them."

Why marginal CAC is unmeasured white space

Here is the part operators should sit with. Blended and new-customer CAC have published benchmark distributions: Shopify's US$127 to US$377 by-industry figures, US$78 blended, vertical-by-vertical tables from Triple Whale's 2025 dataset of 33,000-plus brands and US$18.4B in ad spend, where the overall median paid CPA sits near US$33 (US$32.74 in 2025; US and global DTC data; Triple Whale 2025). Marginal CAC has no published benchmark. Search for "marginal CAC benchmark by vertical" and you will find definitions and pleas to measure it, but no distribution, no medians, no quartiles. The one number that actually governs the scale decision is the one the industry has never benchmarked.

That is not an accident. Marginal CAC cannot be looked up; it has to be computed from your own spend-and-acquisition history, because the slope of the curve is specific to your account, your channel mix, and your current operating point. It requires holding spend and new-customer counts side by side over time and taking the first difference, ideally validated against a real incrementality test rather than platform-reported conversions, which over-report. It is more work than reading a benchmark, which is precisely why most brands quote the average and scale on a number that was never built to answer the question they are asking.

The practical discipline is three lines on one screen. Report blended CAC to judge the whole machine. Report new-customer CAC to compare yourself to peers. And compute marginal CAC, by channel, before every budget change, because that is the only one of the three that tells you whether the next dollar makes money. Get those three numbers separated and labelled, and "what's our CAC" stops being a trick question.