New: results from 5,000 simulated shoppers are in. See the results
Send or Silence
Detailed numbersTables, charts and the per-persona breakdown
Loading results…

How the experiment works

Pretend shoppers with secret habits

Six kinds of shopper: WhatsApp fans, email readers, night owls, bargain hunters, easily annoyed people and window shoppers. Each browses, fills carts and sometimes buys on their own, and reacts to messages in their own way. Too many messages annoy them, and messages at night annoy them more.

A fair race

Every approach gets exactly the same shoppers doing exactly the same things. “Extra sales” means sales on top of what happens when nobody gets any reminders at all. The engine never peeks at the secret habits (a test checks this); it has to learn them from what shoppers do.

No cheating on the numbers

Settings could only be adjusted on practice runs. The results shown come from separate runs done once, afterwards, and they’re reported as-is, including where the engine loses.

Limits: the shoppers are simulated and their habits were chosen by the author; one time zone; every message is delivered; “opened” stands in for “worked”.

Try it yourself

Run a smaller version of the experiment (takes a few seconds). The headline numbers above come from the full, frozen runs.