Blended CAC vs. Channel CAC — Which One Goes to the Board? (both numbers are correct)

Meta Ads Manager says your CAC is ₹842. Google Ads says ₹918. Your finance sheet says ₹1,472. Everyone is looking at the same month, the same spend, the same customers.

The performance marketer trusts the platforms — after all, each dashboard divides its own spend by its own conversions, and the arithmetic checks out. Finance trusts the bank statement. The gap between them is ₹630 per customer, which across 1,600 new customers is more than ₹10,00,000 of unexplained optimism in a single month.

Both numbers are correct. They answer different questions. Channel CAC asks "how efficiently did this channel turn its own spend into the conversions it claims?" Blended CAC asks "how much did we actually spend to create one genuinely new customer?" Only the second one reconciles to money that left the bank.

Why channel CAC always looks better

Channel CAC is a fraction: channel spend divided by channel-claimed conversions. Three structural biases push that fraction down, and every one of them is working at the same time.

Take a concrete month. Meta spent ₹8,40,000 and claimed 998 conversions: CAC of ₹842. Google spent ₹7,20,000 and claimed 784 conversions: CAC of ₹918. Added together, the platforms claim 1,782 new customers from ₹15,60,000 of spend — an apparent combined CAC of ₹875.

The warehouse tells a different story. Deduplicated on email and phone, only 1,180 genuinely new customers arrived that month. Some platform-claimed conversions were the same person counted twice. Some were existing customers clicking a retargeting ad and being re-counted as "new." Some were view-through attributions on people who never noticed the impression. The platforms are not lying; each one is reporting exactly what its own attribution logic saw. But attribution logic is designed to justify spend inside that platform, not to reconcile across them.

Then there is the cost side. The ₹15,60,000 of platform spend sits inside ₹17,40,000 of true acquisition cost once you add the agency retainer apportioned to acquisition (₹90,000), creator and creative costs (₹65,000), and the app-install and SMS incentives tied to first orders (₹25,000). Channel CAC divides a partial cost by an inflated customer count. Both errors point the same way — down.

What blended CAC actually reconciles against

Blended CAC has a deliberately boring definition:

Blended CAC = All acquisition spend ÷ New customers from your own orders table

Boring is the point. Every input comes from a system you control, and every input reconciles to something auditable.

The numerator reconciles against the P&L. Start from total marketing cash outflow for the month, then remove what is genuinely retention: lifecycle email/SMS tool costs aimed at existing customers, retargeting spend against past-purchaser lists, loyalty rewards. What remains is the cost of creating new customers, and it should tie — within a few percent — to the marketing lines in your books. If it does not tie, your definition is leaking.

The denominator reconciles against the orders table. A new customer is someone whose first ever paid order falls in this period, deduplicated on a stable identity like email or phone hash — never on a platform cookie or a customer_id that guest checkout regenerates. A customer who bought in January and again in June is a returning customer in June, even if Google claims them as a fresh conversion because they clicked a brand ad. Your warehouse is the only system that can see this, because it is the only system that sees every channel at once.

Computing both from the warehouse

The discipline that ends the argument is computing both numbers from the same tables, so the gap is visible and its components are named.

The column to watch is claimed_per_real_customer. At 1.0, platforms collectively claim exactly the customers you actually got. In practice it runs 1.3–1.8 for multi-channel D2C, and anything above 1.5 means more than a third of "conversions" are double-counts, returning customers, or view-throughs on people who would have bought anyway. Track it monthly — when it jumps, someone changed an attribution window or launched a retargeting campaign that is harvesting existing demand. 1.5+ means a third of conversions are ghosts

Board reporting: blended. Channel optimization: channel.

The mistake is not using the wrong number. It is using one number for both jobs. They belong in different rooms.

Three rules keep the peace. First, never compare channel CACs across platforms as if they were measured in the same currency — a ₹842 Meta CAC and a ₹918 Google CAC do not mean Meta is cheaper, because each denominator is self-reported under different windows and different view-through rules. Compare trends within a channel, and compare channels only on the bridged new-customer CAC from your warehouse. Second, never set budgets from channel CAC alone, or you will scale the channel that is best at claiming credit — usually brand search and retargeting — and starve the prospecting that actually creates demand. Third, report the gap itself: blended alongside the spend-weighted channel average, with claimed_per_real_customer as the footnote. When the board can see the gap, nobody has to accuse anyone's dashboard of lying.

For Indian D2C specifically, add one more reconciliation step: compute blended CAC on delivered orders, not placed orders. With COD and RTO running 15–35%, a meaningful share of "new customers" never become revenue. Dividing acquisition spend by placed-order customers understates true CAC by the RTO rate — divide by delivered-order new customers instead, and the number finally matches cash.

Quick gut-check

Meta says ₹842, Google says ₹918, finance says ₹1,472: everybody's right, everybody's answering differently!

Frequently asked questions

Why is blended always 30–70% higher than channel CAC?

Because the two gaps compound. Platforms overcount customers by 20–40% through double-claiming and returning-customer capture, and undercount cost by 15–30% by excluding overheads. Multiply a 1.3x customer inflation by a 1.2x cost understatement and you get a 1.56x gap — a ₹900 channel CAC beside a ₹1,400 blended CAC is the expected outcome, not a data bug.

Should agency fees and creative costs really go into CAC?

Yes, apportioned to acquisition. If you fired the agency and stopped producing creatives, acquisition would fall — so they are acquisition costs. Exclude pure retention work (lifecycle management of existing customers), but include everything that exists to create new customers. The test is marginal: would this cost disappear if you stopped acquiring?

Is brand search acquisition or retention?

Mostly retention wearing an acquisition costume. Someone searching your brand name already knows you; the ad harvests intent created earlier by other spend or by word of mouth. Keep brand search spend and its conversions out of prospecting channel CAC comparisons, and count it as acquisition cost in blended only with a clear label so its weight is visible.

How often should we reconcile the two?

Monthly, on a fixed cadence, with the warehouse query above. Daily reconciliation is noise — delivery lags, COD confirmation delays and attribution delays all need days to settle. Monthly is frequent enough to catch a broken window change or a retargeting campaign eating prospecting budget, and slow enough that the numbers have stabilised.

Can blended CAC be gamed?

Less easily, but yes — by widening the definition of "new customer" (counting resurrected guest checkouts as new), by excluding inconvenient overheads from the numerator, or by computing it on placed rather than delivered orders in a COD-heavy business. The defence is the same discipline that makes it trustworthy: publish the numerator components and the denominator definition next to the number, every month.

The summary

• Channel CAC divides partial cost by self-reported conversions. Shared credit, cheap demand capture and missing overheads all push it down.
• Blended CAC divides all acquisition spend by warehouse new customers. It reconciles to the P&L and to the orders table, which is why it is higher — and why it is honest.
• Watch claimed_per_real_customer: 1.3–1.8 is normal for multi-channel D2C, and a jump means a window changed or retargeting is harvesting demand.
• The board gets blended, the media buyer gets channel, and cross-channel allocation gets the warehouse-bridged new-customer CAC. Never compare channel CACs across platforms.
• On COD-heavy catalogues, compute blended on delivered new customers — placed-order CAC understates reality by the RTO rate.