Blended CAC Across Paid Channels: A Reconciliation Guide

Add up the revenue each ad platform reports and the total lands above the revenue the business actually banked. That does not by itself prove a tracking fault — it can happen when multiple channels claim credit for the same purchase, though reconciliation is needed to identify the cause. A companion piece handles how to assign each channel a role and judge the marginal dollar. This guide does the other half: the business-level reconciliation, lining up all-in acquisition spend, net-new customers, actual revenue, fees, and channel-reported attribution on one ledger so no number is counted twice.

For the neighboring economics, compare Owned Channels Are an MER Lever: Shift Spend When CAC Rises and use Paid-Social Checkout Teardown: From Cart to Thank-You to validate the measurement decision.

In short

The decision is reconciliation, not ranking. You want one figure that survives every platform’s self-scoring — blended CAC = total defined acquisition spend ÷ net-new customers — computed on aligned populations, so the spend in the numerator and the customers in the denominator describe the same set of people. Channel-reported revenues stay steering signals reconciled to the order ledger, never summed into business revenue. Marketplace fees and fulfilment are sale costs, not ad spend, and stay out of every advertising denominator.

The reconciliation problem (why the reports do not add up)

Each platform reports conversions on its own attribution window and its own pixel, so one customer can surface in more than one dashboard for the same purchase. Meta may credit a click within its window or a view it served; search may credit the same owned-store purchase to the last click before checkout; a marketplace credits the sale to itself, on its own separate ledger. A buyer who saw a paid social ad, later searched the brand, and checked out on the owned store can appear in full in two owned-store dashboards at once. The reconciliation job is to collapse that back to what the business banked, and to name every place a number is at risk of being double-counted.

Measurement dictionary (numerator ÷ denominator for every metric)

Every metric below states its own math. If a later section cites a number, this is where its definition lives.

• Blended CAC = total defined acquisition spend ÷ net-new customers. Align both populations. The numerator is only the spend that acquires the customers counted in the denominator; the denominator counts a customer as new only on first identified purchase. If a channel’s customers cannot be identified or deduplicated (a marketplace, below), its spend does not belong in this numerator against a denominator that excludes its buyers.
• Net-new customers = distinct customers whose first identified purchase falls in the period. Identified in the order system by a durable identifier (email or account), not by any ad platform. Net-new can never exceed orders in the period.
• MER (marketing efficiency ratio) = total business revenue ÷ total paid-media spend. A whole-business ratio. Do not call this “blended ROAS”: its denominator is paid media only, across every channel at once, not any single channel.
• Channel ROAS = that channel’s attributed revenue ÷ that channel’s ad spend. A per-platform steering measure on that platform’s own ledger. It is not proof of incrementality, and two channels’ ROAS are not directly comparable — different ledgers, different cost stacks.
• Reporting excess (non-additivity gap) = channel-reported revenue in excess of the actual revenue on that ledger. When platforms sharing a ledger (Meta and search on the owned store) report revenues that sum above what the store banked, the excess proves only that the reports cannot be summed — it is non-additivity, not a measured count of how much revenue was duplicated. Attribution overlap (both platforms crediting the same orders) is one scenario explanation for the gap; sizing the true overlap needs order-level deduplication or a test. It is never added into the revenue line.
• Contribution per order = AOV − (cost of goods + shipping + returns + channel fees + fulfilment). What an order leaves behind after its own costs. Referral fees and fulfilment are sale costs and live here — not in any ad-spend denominator.

Reconciliation table (one worked scenario)

The figures below are an assumption, not a benchmark — one internally consistent scenario so you can see how the ledgers relate. Recompute against your own accounts. Two revenue ledgers are shown: the owned store (fed by Meta and search) and the marketplace.

Illustrative scenario — an assumption, not a benchmark or an expected result.

Line	Owned store	Marketplace
Orders	2,400	900
AOV	$80	$65
Actual revenue (= orders × AOV)	$192,000	$58,500
Meta ad spend	$30,000	—
Search ad spend	$12,000	—
Marketplace ad spend	—	$9,000
Agency + tooling (owned-store acquisition)	$6,000	—
Meta-attributed revenue (platform-reported)	$132,000	—
Search-attributed revenue (platform-reported)	$84,000	—
Marketplace-ads-attributed revenue (inside marketplace)	—	$27,000
Net-new customers (identified first purchases)	1,600	unknown

Actual business revenue = $192,000 + $58,500 = $250,500. That is the only revenue the business banked, and the only revenue figure any business-level metric may use.

Now watch the channel-reported totals against actual owned-store revenue. Meta-attributed ($132,000) + search-attributed ($84,000) = $216,000 of channel-reported revenue against $192,000 of actual owned-store revenue. That $24,000 reporting excess (12.5% of owned-store revenue) is the non-additivity gap — evidence the two reports cannot be summed, since it exceeds what the store banked. This scenario treats attribution overlap (both platforms claiming credit for some of the same owned-store orders) as one explanation for that excess; the $24,000 is not a measured count of exactly that much revenue duplicated, and pinning the true overlap needs order-level deduplication or a test. Either way it is not additional business revenue.

Blended CAC on aligned populations (the core discipline)

The one reconciliation mistake that quietly corrupts a blended number is misaligned populations: dividing a numerator that includes one channel’s spend by a denominator that excludes that channel’s customers. So this scenario computes blended CAC on the population it can identify — owned-store first purchasers — paired with only the spend that acquires them.

• Total defined acquisition spend (owned-store-identifiable) = Meta $30,000 + search $12,000 + agency/tooling $6,000 = $48,000. Agency/tooling is a real acquisition cost, so it is included; marketplace ad spend is excluded because its buyers are not in the denominator.
• Net-new customers (denominator) = 1,600 owned-store first purchasers, identified in the order system, and 1,600 ≤ 2,400 owned-store orders (the rest are repeat purchases).
• Blended CAC = $48,000 ÷ 1,600 = $30.00.

The marketplace ad spend sits outside that ratio, reported on its own basis because its customer identity is limited:

• Marketplace ad spend per marketplace order = $9,000 ÷ 900 = $10.00 per marketplace order. This spreads spend across all marketplace ledger orders — not CAC per new customer, because the marketplace net-new count is unknown and cannot be deduplicated against the owned store. Folding $9,000 into the $48,000 numerator while leaving the denominator at 1,600 owned-store buyers would divide one channel’s spend by another population’s customers.

MER and channel ROAS from the same table

The steering measures derive directly from the table, and each keeps its own denominator.

• MER = total business revenue ÷ total paid-media spend = $250,500 ÷ ($30,000 + $12,000 + $9,000) = $250,500 ÷ $51,000 ≈ 4.9×. Total revenue over total paid media — not “blended ROAS” (the denominator is paid media across every channel, not one channel). Agency/tooling is excluded here because MER is defined on paid media, yet included in blended CAC because that ratio is defined on acquisition spend — different, explicitly-stated denominators.
• Meta ROAS = $132,000 ÷ $30,000 = 4.4× (Meta-attributed owned-store revenue ÷ Meta spend).
• Search ROAS = $84,000 ÷ $12,000 = 7.0× (search-attributed owned-store revenue ÷ search spend).
• Marketplace ads ROAS = $27,000 ÷ $9,000 = 3.0× (attributed inside the marketplace ledger).

The 4.4×, 7.0×, and 3.0× cannot be summed and are not directly comparable: the first two share the owned-store ledger (so their attributed revenues are non-additive, per the $24,000 reporting excess — attribution overlap is one explanation for it), and the third sits on a separate marketplace ledger with a fee-and-fulfilment cost stack the others do not carry.

Fees and fulfilment: sale costs, not ad spend

A marketplace order looks efficient on ad spend and less so after its own costs — which is why fees belong in contribution, never in the ROAS or CAC denominator. At 60% gross margin in this scenario, marketplace contribution per order = $65 AOV × 60% = $39 gross contribution, minus a 15% referral fee ($9.75) and $4 fulfilment = $25.25 per order before any ad-spend allocation; subtracting the per-order marketplace ad allocation ($10.00) leaves $15.25 average ledger contribution. Keeping the $9.75 fee and $4 fulfilment out of the ad-spend-only ROAS denominator is what stops that line from looking like a clean 3.0× while its contribution reads lower. The 15% referral figure is a scenario assumption — real marketplace commissions vary by marketplace and product category, so use your own contract rate rather than any single number as universal.

Customer identity limits (do not assume marketplace buyers are net-new)

The denominator of blended CAC lives or dies on identity. On the owned store you can identify a first purchase by a durable identifier and count net-new customers (1,600 here) — the population the $48,000 numerator and the $30.00 CAC describe. On a marketplace you may not receive the buyer’s email or a durable identifier, so you cannot deduplicate a marketplace buyer against owned-store customers: “net-new” is unknown, not “yes”, which is why the marketplace row shows unknown and its ad spend is reported per order rather than folded into blended CAC. A figure is only “blended CAC” if its denominator genuinely includes deduplicated identities; if you truly resolve some marketplace identities, add exactly those to both sides, otherwise keep the two reconciliations separate. A marketplace order is real revenue on the marketplace ledger, of unknown incrementality to your owned-store program, until a controlled test says otherwise.

Running the reconciliation

Pull total defined acquisition spend (all-in, with agency/tooling) and net-new customers from the order system, not a dashboard, and compute blended CAC on aligned populations plus MER, each with its denominator written down. Track channel-reported revenue minus actual revenue on each shared ledger over time: a widening gap is a growing non-additivity gap to reconcile (attribution overlap is one explanation to size with order-level dedup or a test), not revenue to bank. Keep channel ROAS as a steering wheel for moving budget inside a platform, never reported upward as settled truth. And tie CAC to margin — a $30.00 blended CAC is only worth paying if first-order contribution (revenue minus cost of goods, shipping, payment fees, and returns) covers it inside a payback window you can fund. Model contribution and CAC separately.

Incrementality is a separate question (test, do not assume)

Reconciliation tells you what the business banked and what each spend cost against an aligned population. It does not tell you what would have happened without a given channel — that is incrementality, and the only clean read is a controlled change. Hold a defined audience or matched regions out of a channel, or step its spend up or down, and measure the change in total business revenue and orders across both ledgers, reconciled to your order records rather than the platform’s attributed number. State the limits — a holdout needs enough scale and duration to separate signal from noise, seasonality and outside events can confound a single test, and one result is specific to the audience, period, and creative you ran — and re-run when conditions change.

Can software help?

Bach.ai audits your connected Meta account against 100+ checks, ranks what it finds by estimated impact, and proposes specific fixes. It stays read-only until you approve a change, then executes the approved change on Meta; connected Google Ads data is used for intelligence only. Think of it as an automated audit layer that surfaces issues and proposed fixes for your review — not a replacement for your team’s judgment, and it does not generate your creative.

FAQ

How do I calculate blended CAC without double counting?

Total defined acquisition spend ÷ net-new customers, with the two populations aligned. Include the spend that acquires the customers you can identify (here Meta $30,000 + search $12,000 + agency/tooling $6,000 = $48,000) and divide by first purchasers (1,600) for $30.00. Spend for a channel whose buyers you cannot identify (marketplace $9,000) is reported on its own basis ($9,000 ÷ 900 orders = $10.00 per order), not folded into the ratio.

What is the difference between MER and blended ROAS?

MER (marketing efficiency ratio) = total business revenue ÷ total paid-media spend. There is no separate “blended ROAS” here: when the denominator is paid media only, the correct name is MER. Calling it “blended ROAS” implies a single-channel efficiency read it does not provide.

Should marketplace fees count in CAC or ROAS?

No. Referral fees and fulfilment charges are costs of the sale, not advertising cost. Keep them out of any ROAS or CAC denominator (both are spend-based) and account for them in contribution per order instead.

Are marketplace buyers net-new customers?

Treat that as unknown. Marketplaces limit the buyer identity you receive, so you may not be able to deduplicate a marketplace buyer against owned-store customers. Count marketplace orders as revenue on the marketplace ledger and compute blended CAC only on the population you can identify.

How do I know if a channel is actually incremental?

Reconciliation cannot answer that; a controlled test can. Run a holdout or geo test — withhold a defined audience or matched regions, or step spend up and down — and measure the change in total business revenue and orders across both ledgers, reconciled to your order records rather than the platform’s attributed number. State the limits (scale, duration, seasonality, confounds) and re-run when conditions change.