Lookalike Audience

Deciding who sees an ad was, for a long time, a matter of assembling conditions by hand. Cut the age range at 25 and up, then pick "productivity" from the topic labels the platform has attached to people.

The limit of that approach is that those labels are estimates you cannot inspect. Set interest targeting to productivity and the reachable pool comes back at 300,000 — a pool holding both people genuinely shopping for a meeting-notes tool and people who once tapped a related post.

Using your own customer list instead is accurate but capped at the size of the list. Three thousand customers cannot bring in anybody new.

So the method that emerged hands the judgment to the platform. It holds far more behavioral history than you do, so instead of writing conditions you give it a few thousand customers who stayed as examples and ask it to find people like them. What counts as similar is decided by the platform, and the advertiser cannot see the criteria.

Build and upload a seed list, choose how wide the resemblance should reach, then run the resulting audience alongside your other conditions.

1. Choose the seed. Handing over every signup lets people who used the free plan and left define what resemblance means, so pick only customers who have stayed on a paid plan for three months or more. Platforms enforce a minimum, usually somewhere between 1,000 and 2,000 people.
2. Upload it. Email addresses and phone numbers go up in an unreadable form and get matched against the platform's own values. The custom audience produced this way becomes the seed. Which data goes where falls under marketing privacy rules, so check before uploading.
3. Set the width. A top 1 percent audience is narrow and expensive; 10 percent is broad and cheap. Start between 1 and 3 percent and widen based on results.
4. Layer other conditions. Keep the geography and age filters, and add existing customers to an exclusion list.
5. Judge it on cost per signup. A higher cost per click is acceptable if cost per signup falls. Rebuild the seed quarterly, since who your customers are keeps changing.

"We seeded it with every signup and got a flood of free-plan-only users. Let's switch to paid for three months or more."

The quality of the seed is what defines resemblance.

"One percent is too small a pool. Widen it to three so we actually spend the daily budget."

Too narrow to fill the impressions.

"We never set the exclusion list. We're running signup ads at our own customers."

Budget is leaking because the exclusion step was skipped.

At Pagedeck, Ravi seeds this with the 1,800 customers who have stayed on a paid plan three months or longer, at a width of 2 percent. Cost per click went up, while cost per signup fell from $25 to $15.