When the women’s apparel brand M.M.LaFleur began advertising on Facebook in 2013 or 2014, it could acquire a customer for $13 or $14. Last year, founder and CEO Sarah LaFleur told the podcast Odd Lots that the figure was closer to $250, “and that’s probably understating it.”
LaFleur added that Facebook was probably still the company’s most effective advertising channel. She may well be right. Her experience nevertheless raises a question confronting advertisers throughout the economy: As customer acquisition becomes more expensive, what evidence can tell a business where its next marketing dollar will create the most value?
Many advertisers lack a coherent framework for making that decision. Instead, they react to the metrics their platforms and measurement products make readily available. Platform customer acquisition cost (CAC) is especially seductive because it provides a precise, frequently updated number and suggests an obvious action: spend more where CAC is lowest. But the metric at hand is not necessarily the evidence needed to answer the larger business question.
Optimizing CAC is not the same as optimizing marketing’s contribution to the business. Its apparent simplicity conceals six problems.
The easiest customers get acquired first.
Reported CAC is probably not incremental.
Performance measurement creates a vicious cycle.
Platform-reported CAC does not capture the whole market.
Platform CACs are not comparable.
Customer acquisition represents only part of advertising’s total return.
Understanding these limitations points toward a better approach: a coherent measurement framework organized around business outcomes and investment decisions, rather than a collection of piecemeal platform metrics.
The Easy Customers Get Acquired First
When a young company begins advertising, it has nothing but potential new customers. An unfamiliar brand with a distinctive position may find an audience that is highly receptive but has never encountered it before. A growing platform may also offer relatively inexpensive access to those prospects.
Over time, the economics change. The brand acquires many of its most likely customers, competitors pursue the same segment and auction prices rise. Additional spending must reach people who are more expensive or less likely to buy. Even a well-run campaign becomes less efficient as it scales.
This is diminishing returns, and it probably explains part of M.M.LaFleur’s experience. The company and Facebook have both matured since 2013. The first customers were never evidence that the hundred-thousandth customer would cost the same amount to acquire.
Reported CAC Is Probably Not Incremental
Cost per acquisition sounds straightforward: divide advertising spending by the number of customers acquired. But the CAC displayed in campaign reporting is usually an attributed measure. It counts customers credited to advertising under a platform’s or advertiser’s attribution rules, including some who might have purchased anyway.
Incremental CAC asks a harder question: How much did it cost to acquire customers who would not otherwise have been acquired? Answering it rigorously requires a randomized control group to estimate the counterfactual: how many customers would have been acquired without the advertising. Some platforms offer such tests, but they are generally occasional studies, not scalable as a continuous operating metric. It’s not the CAC most advertisers are talking about.
Marginal incremental CAC goes further, asking what it will cost to acquire the next additional customer as spending increases.
These figures can differ substantially. An attractive attributed CAC may reflect a platform’s ability to identify people already likely to buy. Even an experimentally measured average incremental CAC does not reveal what another customer will cost after the advertiser increases spending.
Advertisers that measure and calculate attributed, incremental and marginal CAC must keep them distinct. They answer different questions and cannot be treated as interchangeable inputs to investment models. The CAC advertisers use most frequently may be poorly suited to allocation decisions because it describes who received credit for past conversions, not necessarily what caused them or where the next dollar will work hardest.
Performance Measurement Creates a Vicious Cycle
When customer acquisition costs rise, advertisers often shift more money toward the channels offering the clearest and fastest performance reporting. That response can make the underlying problem worse.
Mutinex CEO Henry Innis has warned against flooding into short-term channels when demand is weakening. More advertisers compete against one another in platform auctions for a limited pool of high-intent customers, bidding up prices just as consumers become harder to convert. Cutting brand investment can also weaken the future demand that performance media harvests.
A feedback loop develops. Performance channels receive more money because their results are easier to measure. Greater spending produces higher auction prices and diminishing returns. Rising CAC creates pressure for still more short-term optimization, while less money is devoted to building future demand.
Closed-loop measurement requires another consequential exchange. Through Meta’s Conversions API and similar systems, advertisers send customer identifiers, purchase events and transaction values to the same companies selling them media. Repeating that process across multiple platforms expands the security, privacy and governance burden surrounding some of the advertiser’s most sensitive data.
Don Marti has argued that these data flows increase Big Tech’s leverage over advertisers. Advertisers supply the information platforms use to measure their own performance, improve their optimization systems and make themselves more difficult to leave. The advertiser assumes the integration and data-governance risks, while the platform accumulates the informational advantage.
Platform-reported CAC Does Not Capture the Whole Market
Platform measurement can observe only the customers and transactions available to it. Meta’s Conversions API can receive offline transactions, for example, but it cannot match purchases for which usable customer identifiers are unavailable. Sales outside direct-to-consumer channels are often difficult to match to advertising exposures. Retailers and other intermediaries may share only aggregated sales totals rather than transaction-level customer identifiers. Wholesalers may report shipments to retailers without revealing the eventual consumer purchase. Cash transactions, purchases made without loyalty accounts and other unidentified sales leave no matchable customer record. Even when a platform’s conversion API can accept offline sales, it can measure only the portion connected to usable identifiers.
The visibility problem also works in the other direction. Television, CTV, radio, out-of-home and publishers may create awareness and preference that later produce searches, site visits or purchases through Meta, Google or Amazon. Those channels often lack equally convenient systems for connecting their exposures to the eventual sale.
Platform-reported CAC therefore reflects the portion of the customer journey that the platform can see and claim. It is not necessarily a comprehensive measure of how the business acquired customers or which marketing investments created the demand. Money can follow measurement capability rather than media effectiveness.
Platform CACs Are Not Comparable
Randomized platform lift studies are a meaningful improvement over attribution. But even a rigorous platform experiment does not necessarily produce a CAC that can be compared directly with another platform’s result.
A platform lift test measures incremental acquisition against that platform’s own audience, delivery system and counterfactual baseline. Placing its iCAC beside another platform’s result does not create a common measure of comparative ROI.
Meta, Google, TikTok and Amazon each reach different users through different optimization systems. Their experiments may be internally valid while estimating effects for different populations under different conditions. A Meta iCAC and a TikTok iCAC can both be accurate without being apples-to-apples.
Comparing them is like building a house using feet for the kitchen, meters for the living room, cubits for the bedrooms and barleycorns for the garage. Every measurement may be accurate, but the measurements cannot be combined until they are placed on a common basis.
Customer Acquisition Represents Only Part of Advertising’s Total Return
New-customer acquisition is essential to growth. But winning a customer once does not secure that customer’s future purchases. As Byron Sharp has emphasized, buyers in most categories regularly move among competing brands rather than remaining exclusively loyal to one.
Advertising competes to influence every next purchase, whether it comes from a first-time customer or a regular buyer. Erwin Ephron’s recency theory emphasized advertising’s role in influencing consumers near the time of purchase. In Sharp’s terms, advertising helps keep the brand mentally available when the next buying occasion arises.
Advertising may cause an existing customer to choose the brand again, purchase sooner or spend more. CAC excludes that value by design. A channel with a relatively high incremental CAC may still generate strong incremental revenue and profit through purchases by existing customers. Conversely, a channel with an attractive attributed CAC may contribute little once customers who would have purchased anyway are removed.
The CFO’s question is not simply how cheaply marketing acquired new customers. It is how much incremental economic value the investment created overall.
The Business Needs a North Star
Advertisers need more than a collection of campaign metrics and isolated lift studies. They need an evidence-based framework for determining how much to invest in marketing, where to allocate it and whether that investment is improving the business.
My friend Talgat Mussin recommends Marketing Efficiency Ratio (MER), total revenue divided by marketing expenditure, as a simple business-level North Star. Unlike platform CAC or ROAS, MER focuses attention on the performance of the whole business rather than the results claimed by individual media sellers. But it is not causal. Revenue also changes because of pricing, distribution, seasonality, product demand and many other factors unrelated to advertising.
That is why no single metric or measurement product is sufficient. Platform reporting can help improve execution within a platform. Independent geo experiments provide a broader basis for comparison by measuring aggregate business outcomes across channels, including transactions that cannot be matched to individual media exposures. Marketing mix models provide the portfolio-wide view, while experiments validate and recalibrate their most consequential estimates over time.
The remedy is not another dashboard. Measurement should begin with the business decision, identify the uncertainty preventing management from making it confidently, and then apply the evidence best suited to resolving it.
The most important question in marketing is not what it cost to acquire the last customer. It is where the next dollar will create the greatest incremental economic value, and what evidence gives the business confidence in that decision.