New vs Returning Customer Acquisition
Date: 2026-08-16
Paying to reacquire people you already had. It inflates apparent new-customer volume, understates CAC, and is the single most common way an ad account looks efficient while adding nothing.
What it is
The distinction between acquiring a genuinely new customer and paying to bring back an existing one — and the fact that most ad platforms report both as conversions without separating them.
platform reports 1,000 conversions
genuinely new customers 400
existing customers, returning 600 ← would many have
returned anyway?
CAC computed on 1,000 is 2.5× better than CAC computed on 400. Same spend, same customers, entirely different conclusion about whether the channel works.
Why it happens by default
Not deception — it’s how the platforms are built:
- Retargeting explicitly targets past visitors, who by definition include past customers
- Broad audiences include your customer base unless you exclude it
- Branded search captures people already coming, most of whom know you because they bought before
- Lookalike audiences are built from customers, so they resemble customers
None of these are wrong as tactics. The error is counting their conversions as acquisition and dividing spend by the total.
The two questions to separate
1. Is this a new customer? Answerable from your own order history, and it’s just a join. The platform can’t tell you — it doesn’t know who bought from you in 2023.
2. Would they have come back anyway? Much harder, and it’s the incrementality question. Retargeting an existing customer who was already returning is spend against a sale you had.
Question 1 is a data problem you can fix this week. Question 2 needs a holdout.
What to do
- Compute CAC on genuinely new customers only. Join platform-reported conversions to your order table and split by first-order flag. This alone corrects most inflated CAC figures — Customer Acquisition Cost
- Exclude your customer list from prospecting campaigns. Basic, frequently not done
- Separate prospecting from retention budgets, and judge them on different metrics. Prospecting on new-customer CAC; retention on incremental repeat revenue
- Run a holdout on retargeting. It’s the highest-value single incrementality test available, because retargeting is where the overlap is worst — Geo Holdout Tests
- Report both numbers, so nobody has to guess which one is being quoted
Retention spend isn’t acquisition spend
The framing that resolves most of this: a returning customer is a retention outcome, not an acquisition one.
That matters because the economics differ completely:
| New customer | Returning customer | |
|---|---|---|
| Cost to reach | High — Customer Acquisition Cost | Low, often owned channels |
| Contribution on first order | Loses money after CAC | Positive immediately |
| Right metric | New-customer CAC, payback | Incremental repeat revenue |
| Right channel | Prospecting | Email, lifecycle — Lifecycle Messaging |
Paying acquisition prices for retention outcomes is the expensive error, and it’s invisible if the two aren’t separated. Email costs pennies to reach a returning customer; retargeting costs pounds to reach the same person.
The measurement dependency
All of this needs reliable identity — knowing whether this person has bought before. Which means it’s downstream of Identity Stitching and of order-system joins, and it degrades exactly where identity does.
A practical floor: use email address matched against the order system, not cookie-based new/returning flags, which decay in days — Browser Privacy Restrictions, User Counting.