Datarithms

How many customers you need before judging a campaign

Datarithms gives no verdict on a campaign until it has at least 30 paying customers in the period. Below that, one customer more or less moves the cost per customer so much that a ranking mostly reflects luck. The campaign is shown as "too few to tell", with its counts.

Last updated:

Why do small numbers swing so much?

With 4 customers, one extra customer is a 25% change. With 40 customers, one extra is a 2.5% change. The cost per customer of a small campaign can halve or double from one week to the next without anything real changing.

One customer more or lessExample data
CampaignSpendPaying customersCost per customerWith one more customerWith one fewer
Job title targeting$1,9604$490$392$653
Lookalike, sign-ups$5,57048$116$114$119

The small campaign's cost moves by up to $163 on a single customer. The large one moves by $3.

Why 30 and not another number?

30 is Datarithms' own rule, not a statistical law. We chose it because at 30 customers one customer more or less moves cost per customer by only about 3%, and because it is easy to remember and check. A campaign with 31 customers is not proven good; it has simply earned a position in the table.

What should you do with a campaign below 30?

  • Decide in advance how much you are willing to spend before judging it.
  • Let it run long enough to collect 30 paying customers, or stop it for reasons other than its cost per customer.
  • Do not compare its cost per customer with established campaigns yet.

Want your own table?

Tell us which ad platforms you use and roughly what you spend. We reply by email within two working days.

Early access: we prepare your weekly table with you from your own exports while the direct connections are built.

We use these details only to reply to you and to decide who to onboard. They are stored on our server in the United States. See Privacy, Data handling and, for Türkiye, the KVKK notice.