Did your campaign cause the purchases?
Run a free incrementality test on your own list
Purchases went up after the campaign. Would they have gone up anyway? Drop in a customer file with who got the campaign and who was held out at random. See how much of the result the campaign caused, with the range it could plausibly fall in.
If you held out a random group, the free incrementality test tool counts how many customers your campaign caused beyond those who would have converted anyway.
Three questions this answers
- Did the customers who got your campaign buy more than the group you held back?
- How many of the purchases did the campaign cause, and what is the likely range?
- Is the difference more than chance, even when you test more than one campaign at once?
What you get back
- How many buyers the campaign gets credit for, how many would have bought anyway, and how many it caused, with the range.
- Each group's purchase rate, with its likely range.
- The lift over the held-back group, in points and as a relative change, each with its range.
- Whether each difference holds up after adjusting for testing more than one campaign.
- If your file has spend: the extra spend per customer and per 1,000 customers reached, with its range.
- How many customers each group needed, and the smallest lift the test could detect next to the lift it found.
- Checks on the test itself: the size of the holdout, enough buyers in every group, enough customers, very large purchases, and rows left out.
- The method named and the R code.
What your file needs
- One row per customer.
- A group column: who got the campaign and who was held back at random. Call the held-back group "control", "holdout" or "no e-mail"; otherwise the tool compares against your largest group.
- An outcome column: bought or not, as 0/1 or yes/no.
- Optional: a spend column, a number per customer with 0 for no purchase, to see the extra spend.
- At least 20 rows, 5 or more in each group. Up to four groups, for example two e-mails and a control.