Glossary - Acquisition and attribution

Lookalike audience

A lookalike audience is a group of new prospects that an ad platform finds for you because they resemble a seed list you upload, usually a list of existing customers. The platform studies the seed, looks for people with similar signals, and serves ads to them. An ecommerce operator uses a lookalike audience to reach cold buyers who behave like the buyers they already have, so the quality of the seed decides the quality of the audience.

Lookalike margin return = (New customers acquired × first-order contribution margin) ÷ Spend on the lookalike

Variable	What it covers
Seed list	The customers you upload for the platform to match against. Not a term in the sum, but the input that shapes every term in it.
New customers acquired	First-time buyers from the lookalike, counted in your store, not the platform's claimed conversions. Returning customers who happened to see the ad do not count.
First-order contribution margin	Revenue less landed COGS, discounts, shipping and payment fees on those first orders, the CM1 basis.
Spend on the lookalike	Media spend on the ad sets targeting that audience over the same window.

There is no formula for the audience itself; the platform builds it and does not show its working. What you can measure is what it returns, so judge it on margin, not on reach or platform ROAS.

Worked example

Two lookalikes, same creative, same budget, same 30 daysA$6,000 each

Seed A (all purchasers): new customers acquired150

Seed A: CAC, A$6,000 ÷ 150A$40

Seed A: first-order margin, 150 × A$28A$4,200

= Seed A margin return, A$4,200 ÷ A$6,0000.70

Seed B (top customers by margin LTV): new customers acquired120

Seed B: CAC, A$6,000 ÷ 120A$50

Seed B: first-order margin, 120 × A$46A$5,520

= Seed B margin return, A$5,520 ÷ A$6,0000.92

Worked example / demonstrative numbers. Seed A wins on new-customer CAC and loses on margin. Neither pays back on the first order alone, so the repeat behaviour of each group decides the verdict, which is why the seed should be chosen on margin LTV.

What is a good lookalike audience?

A good lookalike is one whose new customers earn back their acquisition cost in contribution margin within a payback window you can fund. There is no universal CAC or audience size that marks a good one, because both depend on your margins, price point and repeat rate. What reliably separates good from poor is the seed. A seed of every purchaser teaches the platform to find people like your average buyer, including the one-order discount hunters. A seed built from your highest-value customers, ranked on margin rather than revenue, teaches it to find people like the customers who actually fund the business. Value concentration usually means that group is small, and that is the point.

The second test is whether the lookalike brings genuinely new buyers. Platforms will happily credit a lookalike with purchases from people who were already on their way to buy. A holdout test or geo-lift test answers that; the in-platform report cannot. For the wider question of finding more of your best buyers, see how to find more of your best customers.

Lookalike audience vs related targeting

Approach	Who it reaches	How it differs from a lookalike
Custom audience	People already on your list or site.	Retargets known people. A lookalike uses the same list only as a pattern to find strangers.
Interest targeting	People the platform tags with chosen interests.	You pick the traits. With a lookalike the platform infers them from your seed.
Broad targeting	Anyone in the location and age range.	No seed at all: the algorithm learns from conversions as they arrive.
RFM segmentation	Your own customers, grouped by behaviour.	An internal grouping. The best RFM segment is often the best seed for a lookalike.
Customer lifetime value	A value per customer, not an audience.	The ranking you sort the seed by before you upload it.

Common mistakes

Seeding with every purchaser. The platform matches the average, and the average includes buyers who never came back or only bought on discount. See customers who look like VIPs but lose money.
Ranking the seed on revenue. A big-basket buyer who only buys discounted looks valuable on revenue and poor on margin. Rank on margin LTV instead.
Judging the audience on platform ROAS. The platform grades its own work over its own window. Count first-time buyers in your store and price them in contribution margin, as platform over-claiming explains.
Never refreshing the seed. Your best customers change as the range and pricing change. A seed uploaded once and left alone drifts away from who you want now.
Ignoring overlap. Several lookalikes built from similar seeds compete for the same people and inflate frequency. Check overlap before scaling spend across them.

Lookalike audience FAQ

Your highest-value existing customers ranked on margin, not revenue: people who have ordered more than once at healthy margins. A smaller, sharper seed usually beats a large list of all buyers, because the platform copies whatever pattern you give it, including the unprofitable one.

Count the first-time buyers it brings in your own store data, price their first orders in contribution margin, and compare that with the spend. Then watch whether those customers repeat. To prove the audience caused the sales, run a holdout rather than trusting the platform report.

Often yes, as a test against broad. Broad targeting leans on the algorithm's conversion signal; a lookalike leans on your seed. Run both with the same creative and budget, then compare them on margin return and on how many buyers are genuinely new, not on reported ROAS.

A custom audience is the list itself: people you already know, reached again. A lookalike uses that list as a pattern and reaches strangers who resemble it. One is retargeting, the other is prospecting.

Margin LTV RFM segmentation Value concentration New-customer CAC Incrementality Customer equity Margin LTV calculator All 14 free tools →

Inside Blufire, S8 Persona Analytics shows where each audience over- and under-indexes and the lookalike headroom left, and S9 Activation Bridge pushes an audience to Klaviyo, Google or Meta, confirms it landed, then reads its payback ledger, treatment against holdout.

Updated September 2026