Incrementality case study

Every attribution number has the same hole in it. Some of those customers were going to buy whether or not anyone texted them. We wanted the real figure, so we went looking for a control group, and found one we did not have to build.

WAWashington
OKOklahoma
MDMaryland
FLFlorida

The control group four legislatures built for us

Washington, Oklahoma, Maryland and Florida have messaging laws strict enough that we block texting to them in code. Those customers still see the ads, the site and the emails. The only thing they do not get is the message, so whatever they buy is what would have happened anyway.

A pet grooming brand

Repeat orders

80%p < 0.0001of first-week repeat orders were orders we caused
0%1%2%not textedtexted6 hr1 day2 days3 days7 days14 days

~370

orders a month we caused, this lane alone

0.31% → 1.54%

seven-day reorder rate, untexted then texted

This brand sells a device, and this lane offers the consumable that goes with it. A refill is not something a customer goes looking for in their first week with a new machine, which is exactly why this lane is worth running: almost every refill order here is one the conversation created rather than one we intercepted on its way through. Measured over seven days, because that is where the caused orders stop accruing. Stopping at a day would have counted 142 of them instead of 172.

15,905 customers captured over two weeks. Window: 7 days.

Numbers that survive the next question

Most tools report attribution and hope nobody asks what would have happened anyway. Ask us.

Partner with us