CPA vs CAC vs CPP: stop mixing them in one deck (three acronyms, three jobs)
Published September 16, 2026
The 60-second version
- CPA counts actions, CAC counts new customers, CPP counts platform claims — same spend, three denominators, three jobs.
- True CAC is lead-CPA divided by close rate. ₹420 at 35% close is really ₹1,200 — budgets priced on the first buy fiction.
- The fix: board gets blended CAC, buyer gets channel, and every column names its denominator.
Marketing reports a ₹420 CPA. Finance reports a ₹1,150 CAC. The founder quotes a ₹980 cost per purchase from the Meta dashboard. Same month, same spend, three numbers — and all three are correct.
The confusion is not arithmetic. It is that CPA, CAC and CPP count different people over different money, and each answers a different question. Mix them in one deck and every budget conversation misfires.
The three numbers, precisely
One sentence each, then the maths. A lead is a hand raised. A customer is money in the bank. A platform purchase is whatever the platform decided to claim.
CPA — cost per action
Spend divided by counted actions. The action can be a lead, a trial, a purchase — whatever the campaign optimised toward. Answers: how efficiently did this campaign buy its chosen event?
CAC — cost per customer
All acquisition spend divided by new customers in your warehouse. Blended CAC uses every rupee; channel CAC uses one channel's spend. Answers: what did a new customer truly cost the business?
CPP — platform cost per purchase
One platform's spend divided by that platform's self-reported purchases. Answers: what does the platform think its purchases cost — inside its own counting rules?
| Numerator | Denominator | Source of truth | |
|---|---|---|---|
| CPA (purchase) | Campaign spend | Counted purchase events | Ad platform or warehouse, must be labelled |
| Blended CAC | All acquisition spend | New customers (warehouse) | Your orders table — the board number |
| Channel CAC | One channel's spend | That channel's claimed new customers | Platform — the buyer number |
| CPP (Meta/Google) | That platform's spend | That platform's claimed purchases | Platform — includes repeats, views, modelled |
Two distinctions do all the work. CPA counts actions (one person can convert twice); CAC counts new customers (one person counts once, ever). And warehouse numbers count money once; platform numbers each count the same sale separately.
CPP is not CAC, even when both say per purchase. CPP divides platform spend by platform-claimed purchases — repeats included, returns ignored, views and modelled conversions inside. CAC divides your spend by your new customers — deduplicated, refund-aware, counted once. CPP will always look cheaper. That is the point of CPP.
Why mixing them breaks decisions
Three real misfires, all from one swapped acronym:
Misfire 1: CPA sets the budget, CAC pays the bill. A lead-gen campaign reports ₹420 CPA per qualified lead. At a 35% lead-to-close rate the true cost per new customer is ₹420 ÷ 0.35 = ₹1,200. Funded on ₹420, the plan assumes 2.8x more customers than the money buys.
Misfire 2: CPP judges the channel, blended judges the business. Meta reports ₹980 CPP; blended CAC is ₹1,150 because search, Google and overheads exist too. Kill every channel above ₹1,000 CPP and you kill prospecting that feeds retargeting — total orders fall while every surviving CPP looks prettier.
Misfire 3: purchase-CPA hides repeat buyers. A campaign with heavy returning-customer mix reports a sparkling ₹640 purchase-CPA. New-customer CAC is ₹1,480 because most purchases were not new customers at all. The brand scales retention spend believing it buys acquisition.
📊 ₹420 CPA vs ₹980 CPP vs ₹1,150 CAC
Three fractions, three denominators, one slide. The cheapest number wins the argument — and it is always the platform's narrowest count.
- Budget priced on ₹420, billed on ₹1,150
- Retargeting scaled, prospecting starved
🗂️ CPA optimises, CAC governs
CPA tunes campaigns, CPP tunes platforms, blended CAC governs budgets. Board gets blended, buyer gets channel — never the reverse.
- Targets set on blended payback
- Buyers optimised on labelled CPA
A number without its denominator is a rumour. Label every cost metric with what it counted.
A worked example with one month of data
One D2C month, all numbers reconciled. Watch how the same ₹9,60,000 of spend produces four defensible numbers — and only one governs the budget.
One month, four cost numbers
The ladder reads bottom-up for decisions: blended CAC (₹1,150) sets the budget and the payback clock. Channel CAC (₹1,020) tunes the buyer inside Meta. Purchase-CPA (₹1,066) watches order efficiency including repeats. Lead-CPA (₹420) is a campaign dial, never a business number.
Never average the three. A blended-CAC-meets-CPP average manufactures a number with no denominator at all. Report all three side by side so the size of each adjustment — close rate, repeats, overheads — stays visible.
Computing all three from your warehouse
The structural fix: compute every cost number from one orders table and one spend table, with the denominator explicit in the column name. No platform export decides what a customer is.
One orders table, honest denominators
Run it weekly. Three checks end the confusion: blended CAC is the only series the board sees; every CPA column carries its event in the name (lead_cpa, purchase_cpa, never bare CPA); and the claimed-per-real ratio is the smoke alarm — a sudden jump means someone changed a window or retargeting started harvesting.
Acquisition eats more than one order of margin. Works only with fast repeat — verify with a 90-day cohort before scaling.
Say what to ignore: ignore bare CPA in business reviews, cross-channel CPP comparisons, and any CAC computed on platform-claimed customers. One rule covers all three — board gets blended, buyer gets channel, and every column names its denominator. drag me — your close rate changes everything!
Monday cost-metric hygiene label the denominator →
- Name the event in every column — lead_cpa, purchase_cpa, new_customer_cac. Bare CPA is banned from decks.
- Board sees blended CAC only — channel numbers stay in buyer docs and platform reviews.
- Watch claimed-per-real weekly — 1.3–1.8 is normal; a jump means a window or mix shift, not efficiency.
Quick gut-check
One question. Get it and the whole post clicks. 30 seconds, no maths!
A campaign reports ₹420 lead-CPA at a 35% lead-to-close rate. What budget number should the business plan on?
Frequently asked questions
Is CPP useless then?
No — it is a platform dial. CPP is the fastest signal for intra-platform work: which ad set, creative or audience the platform delivers cheapest. Useless for budgets, excellent for buying. Keep it in buyer docs, out of board decks.
Should repeat purchases count in CAC?
Never. CAC prices the acquisition of a new customer; repeats are retention output. Counting repeats in CAC lets retention subsidise acquisition and hides a dying top of funnel behind a healthy reorder rate. Report repeat revenue separately — cohort curves, not CAC.
Which close rate should divide lead-CPA?
The observed, bounded one: leads from 90+ days ago, closed-won over total, by source. Not the CRM's lifetime hope rate, not the best month's rate. Source-specific, because Meta leads and search leads close nothing alike — one blended rate misprices both.
Why is blended CAC always higher than channel CAC?
Because the numerator is complete (every rupee, including tools, fees and overheads) and the denominator is deduplicated (each new customer counted once). Channel CAC uses partial spend over self-reported claims. Higher is not worse — it is complete. The gap between them is the double-counting tax.
Can I compare my CAC to a benchmark?
Only inside your margin and payback frame. A ₹1,200 CAC is fatal at ₹400 contribution with no repeat, comfortable at ₹900 contribution with 60% 90-day repeat. Benchmark the payback (months to recover CAC from contribution), never the rupee figure alone.
The summary
- CPA counts actions, CAC counts new customers, CPP counts platform claims. Same spend, three denominators, three jobs.
- Lead-CPA divided by close rate is the true CAC. At 35% close, ₹420 is really ₹1,200 — budgets priced on the first buy 2.8x fewer customers than planned.
- Blended CAC (all spend over warehouse new customers) governs budgets. Channel CAC and CPP tune buying inside one platform.
- Purchase-CPA sits between: every paid order counts, repeats included. Useful for order efficiency, never for acquisition payback.
- Compute all three from one orders table with explicit column names. Watch claimed-per-real (1.3–1.8 normal) as the smoke alarm.
- Ban bare CPA from decks. Every cost column names its denominator, and the board sees blended CAC only.
Takeaways for your next report
- CPA counts actions, CAC counts new customers, CPP counts platform claims — three denominators, three jobs.
- True CAC is lead-CPA divided by close rate: ₹420 at 35% close is really ₹1,200.
- Blended CAC governs budgets; channel CAC and CPP tune buying inside one platform.
- Name the denominator in every column and show the board blended CAC only.
Blended CAC Calculator
Total spend across every channel, divided by total new customers — the acquisition cost number that reconciles with what you actually spent.
Chinmay Raibagkar
About author →Founder of DataLens AI. He helps non-technical teams read their ad and database numbers with confidence — which number to trust, what to do next, and what to ignore.