पाठशाला Pathshala · हिसाब Hisāb, Unit economics · Lesson 13 · Build

Blended CAC lies: the channel-level truth

The blended number divides all your spend by all your customers, the free ones included. Split it by channel and by cohort and you find the one channel that pays and the two that hide inside the average.

Pathshala, The Founder Library · 11 October 2026 · 8 min read

Divide a month of marketing spend by the customers who arrived that month and you get a number that looks like the cost of a customer. It is not. It is an average of several businesses, one of which costs nothing, and it is usually low enough to justify spending more on exactly the channels that should get less.

This lesson takes the blended number apart. It works one month of a consumer brand through four channels, defines the three cuts that replace the average, shows where attribution lies and how to test it, and ends with the monthly table that moves budget to where it earns.

What one number hides

Blended CAC counts every new customer in the denominator: the ones won by advertising, the ones who came through a friend, the ones who searched for the brand by name. The organic customers cost the marketing budget nothing, so every one of them pulls the average down. Andreessen Horowitz’s 16 Startup Metrics says the blended figure is not wrong but that it does not tell a founder whether paid campaigns work or are profitable, and that investors weigh paid CAC above it. Bill Gurley made the same point more bluntly in The Dangerous Seduction of the Lifetime Value Formula: if you have organic customers they should not be included in the spend calculus.

The organic line is only the first distortion. The second is that the paid channels are themselves averaged. A good search campaign and a poor influencer programme produce one paid CAC between them, and the good one subsidises the poor one in every report the board sees. The third is time. A channel that was cheap at ₹2 lakh a month is not cheap at ₹12 lakh, and a monthly average blends the cheap first rupees with the dear last ones. Each distortion makes the number lower than the truth for the decision being made, which is where to put the next rupee.

One month, four channels, worked

A skincare brand in Kochi spent ₹22 lakh on acquisition last month and won 4,000 first-time customers. Blended CAC is ₹550. Each first order leaves ₹350 of [contribution](/library/contribution-margin-first-number-to-know) after goods, courier, gateway and returns, and a customer who reorders inside ninety days leaves another ₹400. The founders allow ninety days for a customer to repay what it cost to win them. Across the base 22 per cent reorder in that window, so the average customer returns about ₹440 in ninety days. Against ₹550 that looks close to fine, a gap a little better targeting will close.

Now split it. Meta took ₹12 lakh and won 1,500 customers: ₹800 each. Google search took ₹4 lakh and won 800: ₹500 each. Influencer posts took ₹6 lakh and won 300: ₹2,000 each. The remaining 1,400 customers came direct, through referrals or by searching for the brand name. Paid CAC, with those 1,400 left out, is ₹846 and not ₹550.

The ninety-day repeat rate differs by channel too, because people who search for a remedy behave differently from people who were interrupted by a reel. Search customers reorder at 35 per cent in ninety days, Meta customers at 18 per cent and influencer customers at 10 per cent. So a search customer returns ₹490 against a CAC of ₹500 and almost pays back inside the window. A Meta customer returns ₹422 against ₹800. An influencer customer returns ₹390 against ₹2,000. One channel works. Two lose money on every customer, and they take 82 per cent of the budget. None of this was visible in ₹550.

Paid, channel and marginal CAC

Replace the blended number with three cuts, each answering a different question. Paid CAC is paid spend over customers from paid channels; it answers whether advertising as a whole pays. Channel CAC is one channel’s spend, including agency fees, creative production and the free product sent to creators, over the first-time customers that channel won; it answers where the next rupee should go. Marginal CAC is the extra spend in a channel this month over the extra customers it produced against last month; it answers whether the channel can take more.

Marginal CAC is the one founders skip and the one that matters most when scaling. The Kochi brand raised Meta spend from ₹9 lakh to ₹12 lakh last month and customers from Meta rose from 1,250 to 1,500. The average Meta CAC went from ₹720 to ₹800, a rise nobody would panic over. The last ₹3 lakh bought 250 customers at ₹1,200 each, nearly three times what they repay in ninety days.

The figure opens on the Kochi month. Move the organic slider to zero and watch the blended bar jump to meet paid CAC; that is the distance the free customers were covering. Then move Meta and influencer spend down and see the share of budget above the ceiling fall while the customer count barely moves. Set the ceiling to your own ninety-day contribution, or whatever window your cash allows, and enter last month.

Cut by cohort, because channels age

A channel’s CAC is not a property of the channel. It is a property of the channel at a given spend, a given audience and a given moment. Andrew Chen’s Law of Shitty Clickthroughs records the first banner ad, in 1994, drawing a clickthrough rate of 78 per cent; by 2011 a Facebook banner drew about 0.05 per cent. Audiences tire of a format, competitors copy what works and scale pushes a campaign toward people less likely to buy. The same decay happens inside one company over months.

So track each channel as a series of monthly cohorts: the customers it won in March, in April, in May, each with its own CAC and its own repeat curve. Two things show up that a monthly average hides. A channel’s CAC drifting up across cohorts while spend rises means it is near its ceiling, which the [paid acquisition lesson](/library/paid-acquisition-meta-google-and-the-ceiling) treats in full. And a channel whose later cohorts repeat less than its early ones is reaching a different kind of buyer, which no CAC figure will show. The [cohort lesson](/library/cohort-analysis-for-founders-not-analysts) shows how to build the grid from an order export.

Attribution, and the test that settles it

Every channel CAC rests on a claim about which channel won which customer, and that claim comes from a dashboard that the channel itself often runs. Ad platforms count a customer who saw an ad and later bought, even if the customer was already on the way. Influencer programmes rarely get credit through a tracked link, because people see a post on one phone and search the brand name later on another. So paid platforms tend to over-claim and some channels under-claim, and the organic line is partly made of customers the paid channels created.

Two cheap checks narrow the gap. Ask every first-time buyer at checkout one optional question, where they first heard of the brand, and compare the answers with the platform’s claims each month. Then run a holdout. Switch a channel off for three or four weeks in a set of cities or pincodes chosen in advance, keep it running in a matched set, and compare first-time orders in the two groups including the organic ones. If the influencer programme is feeding the organic line, organic orders will fall where it was paused. If half of Kochi’s organic customers were in fact created by influencers, their CAC would be ₹600 rather than ₹2,000: still above the ceiling, but no longer absurd. Run the test before cutting, not after.

The blended number tells you what a customer cost on average. The channel number tells you where the next rupee should go. Only the second is a decision.

Fix, cap or cut

With channel CAC, channel repeat and a holdout result in hand, each channel falls into one of three cases. A channel under the ceiling with a flat marginal CAC gets more budget, in steps of a fifth a month, with marginal CAC checked after each step. A channel above the ceiling whose holdout showed real lift gets a fix with a date: a new creative, a narrower audience, a different offer, a lower price per creator. A channel above the ceiling with no measured lift is cut, and the money moves to the channel that works.

The gap between channels is often not small. David Skok’s SaaS Metrics 2.0 quotes HubSpot finding a lifetime-value-to-CAC ratio of 1.5 selling direct and 5 selling through partners, the same product sold two ways. A founder looking at the blended ratio would have seen something in between and concluded both were fine. Remember also that cutting the worst channel raises blended CAC for a month or two if the organic line shifts. That is the number becoming honest, not the business getting worse.

The monthly channel table

On the third working day of each month build one table with a row per channel and these columns: spend including fees and production, first-time customers, channel CAC, change in spend and customers against last month, marginal CAC, ninety-day contribution per customer from the cohort three months back, and the ratio of the two. Add a row for organic and direct with customers only, and a last row for blended, so the board sees how far it sits from the rest.

Then make three decisions and write them under the table. Which channel gets more next month, and how much. Which channel gets a fix and by when the fix must show in its CAC. Which channel is cut, or which holdout is run to decide. Once a quarter, compare the checkout survey with the platforms’ claims and adjust the attribution rule. The whole review takes an hour. It replaces a number that flatters with a table that decides.

Nothing here is legal, tax or investment advice. The Kochi brand is illustrative; your ceiling comes from your own contribution and your own cash.

Sources

1. Jeff Jordan, Anu Hariharan, Frank Chen and Preethi Kasireddy, 16 Startup Metrics, Andreessen Horowitz, August 2015 — Blended CAC is not wrong but does not show whether paid campaigns are profitable; paid CAC matters more.
2. Bill Gurley, The Dangerous Seduction of the Lifetime Value (LTV) Formula, Above the Crowd, September 2012 — Organic customers should not be included in the spend calculus.
3. Andrew Chen, The Law of Shitty Clickthroughs — Banner clickthrough of 78 per cent in 1994 against about 0.05 per cent on Facebook in 2011.
4. David Skok, SaaS Metrics 2.0: A Guide to Measuring and Improving What Matters, forEntrepreneurs, January 2013 — Different lead sources carry different costs; HubSpot’s LTV to CAC of 1.5 direct against 5 through partners.