Why customer acquisition cost hides more than it reveals in retail

A blended CACCACCustomer Acquisition Cost (CAC) is the total sales and marketing spend divided by the number of new customers gained in a period. It measures how efficiently you grow.View full definition → (customer acquisition cost) of $28 is an average of four businesses with different economics, printed as one number. A mid-size apparel retailer spends $2 million a quarter across paid social, paid search and in-store window signage. Finance reports the $28, budgets stay flat. Six months later paid social has doubled its cost per customer while search stayed cheap, and nobody caught it, because the average absorbed the damage. And even with a clean mix, $28 tells you nothing about whether this retailer can afford $28.

What CAC actually measures

CAC is the total spend required to acquire one new paying customer over a defined period.

Formula: CAC = Total acquisition spend ÷ Number of new customers acquired

Spend $500,000 in a month, gain 20,000 new customers, CAC = $25.

Two words do most of the hiding. "Spend": does it include the 20% welcome discount that closed the sale? Most retailers book that as a revenue deduction, so it never reaches the numerator and reported CAC understates the real cost by the value of the offer. "New": a lapsed buyer who comes back after 18 months under a different email gets counted as new, padding the denominator and flattering the ratio.

Blended CAC vs. channel-level CAC

Blended CAC pools every channel's spend and every new customer into one ratio. It's the board-deck number: easy to compute, easy to trend.

Channel-level CAC isolates spend and attributed customers per channel: paid social, paid search, in-store signage, email, affiliate.

Take our apparel retailer's quarter:

Paid social is at $50, nearly double the blend, while search and signage are efficient. Watch only the blend and you keep funding paid social because "CAC is fine overall", while two channels subsidise one that is bleeding.

Blended CAC can stay flat or improve while your most expensive channel deteriorates, as long as cheaper channels grow fast enough to offset it.

Why retail specifically struggles with this

• Digital channels have click-level attributionattributionA framework for assigning credit to the touchpoints that contributed to a conversion, so you can measure which channels and interactions actually drive results.View full definition → inside the ad platforms, but attribution windows and last-click logic make each platform the hero of its own report. Meta, which sells the inventory it measures, told investors in February 2022 that Apple's App Tracking Transparency rules would cost it around $10 billion that year. The same signal loss degraded every retailer's view of which customers paid social actually delivered.
• In-store signage, print circulars and local radio have no native attribution. Impact gets inferred from footfall lift, loyalty sign-ups near the store or promo code redemption, all noisy proxies.
• Cross-channel halo is real. Someone scrolls past a social ad, ignores it, then converts after an in-store sign. Last-touch attribution over-rewards the final nudge and under-rewards whatever built awareness weeks earlier.

Channel-level CAC is directionally useful and rarely precise. Treating it as gospel is its own trap.

A cleaner way to compute it: incremental CAC

Ask what happens to new customer volume if you switch a channel off. That is incremental CAC, measured through holdout or geogeoThe practice of making your brand and content visible and citable inside AI-generated answers from tools like ChatGPT, Gemini and Perplexity.View full definition → lift tests: pause spend in a matched set of stores or regions and compare acquisition against a control.

The gap between average and incremental is where budget dies. Auctions reprice as you scale, so the last $100,000 of paid social buys lower-intent audiences than the first, and a channel averaging $50 can carry a marginal cost approaching double that at the top of the spend curve. No ratio built on totals will show you this. Lift studies cost more to set up than pulling a number from a spreadsheet, and they answer the real question: additive, or claiming credit for customers who were coming anyway.

A good primer on incrementalityincrementalityThe share of results (sales, conversions, revenue) that only happened because of a marketing action, not what would have occurred anyway.View full definition → testing methodology is available from Meta's own marketing science documentation, and independent explainers exist on sites like the Interactive Advertising Bureau (IAB), the trade body that sets US digital ad measurement standards.

When a healthy-looking CAC is still unaffordable

Run the $50 paid social customer through unit economics instead of against a benchmark. AOV $120 at 55% gross margingross marginGross margin is the share of revenue left after subtracting the direct cost of producing goods or services, expressed as a percentage of revenue.View full definition → gives $66. Take out pick, pack, delivery, payment fees and returns handling at roughly $18 and contribution is $48. The first order loses money before any overhead.

Returns sharpen it. Online apparel return rates of 20% to 30% are ordinary. If 25 of every 100 acquired customers send everything back and never buy again, your $5,000 bought 75 usable customers: an effective CAC of $67, not $50. The reported figure counted the returner as an acquisition.

Casper, which sells mattresses direct, showed this shape at scale in its 2020 IPOIPOThe first sale of a private company's shares to public investors on a stock exchange, converting private ownership into publicly traded stock.View full definition → filing: sales and marketing running at roughly a third of revenue, net losses of about $90 million on around $440 million of 2019 revenue. No amount of bid tuning fixes that; it needed higher margin, more repeat purchase or less paid dependence. Casper went private in 2022. Gymshark took the other route, building volume through athlete and YouTube influencer partnerships before paid became a main engine, and reached a valuation near £1 billion when General Atlantic took a minority stake in 2020. Cheap acquisition there came from brand demand, not bidding skill.

Whether $50 is survivable depends on the lifetime valuelifetime valueLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.View full definition → the sibling LTV lesson calculates for your category. CAC alone cannot answer it.

Benchmarks: handle with care

Published CAC benchmarks vary by subsector, and most public figures are agency or vendor estimates rather than audited data. Directional, as of early 2025:

• US e-commerce and DTC apparel: blended CAC commonly cited between $30 and $70, including in merchant benchmarking from Shopify, which sells the commerce platform producing the data.
• Grocery and mass-market retail: far lower per customer (single digits to low tens of dollars), because scale and repeat visits spread acquisition spend across huge bases.
• Europe: similar relative patterns, with absolute euro figures published less consistently and varying by national ad market.

A simple diagnostic: the CAC variance check

Before trusting a blended CAC number, run this gut check each reporting period:

For each channel:
  channel_CAC = channel_spend / channel_attributed_customers

blended_CAC = total_spend / total_customers

variance = max(channel_CAC) - min(channel_CAC)

If variance > 30% of blended_CAC:
  flag: "blended number is masking a channel problem"

Applied to the apparel example: variance is $50 − $20 = $30, over 100% of the $28 blend. Loud flag. Any retailer whose channel spread exceeds roughly a third of the blended figure should stop reporting blended CAC as the headline until leadership sees the breakdown.

The reporting habit that fixes this

Not a fancier formula, a reporting discipline: show channel-level CAC alongside blended CAC, never instead of it. Once a quarter, push discount value into the numerator and strip reactivated buyers out of the denominator, even if no platform will do it for you. Boards and CMOs (Chief Marketing Officers) should ask which channel's CAC moved most, and whether customer quality behind it (repeat rate, AOV, return rate) justifies the trend.

🎬 [VIDEO: "Customer Acquisition Cost Explained" - youtube.com/results?search_query=customer+acquisition+cost+explained+retail - search for recent explainer videos from marketing analytics educators covering CAC calculation and channel attribution basics]

Key Takeaways

• Blended CAC can look stable while one channel's true cost per customer deteriorates, because cheaper channels mask it in the average.
• Compute channel-level CAC separately and compare the spread with the blend; variance above roughly 30% of blended CAC means investigate before adding budget.
• Welcome discounts belong in the numerator; reactivated buyers do not belong in the denominator. Both distortions push reported CAC below the real one.
• Attribution is messy: digital over-claims via last-touch, offline lacks native tracking, and post-ATT signal loss made both worse. Holdout and geo-lift tests read truer, and marginal CAC at the top of the spend curve sits well above the average.
• A $50 CAC against $48 of first-order contribution is structurally unaffordable whatever the benchmark says, and apparel return rates of 20% to 30% lift effective CAC by a third when returners never come back.