The welcome-DM experiment: a bot that cut first-week ghosting

The welcome-DM experiment: a bot that cut first-week ghosting

An open-source community ran an A/B test in 2025 across ~6,000 new Discord joiners over three months: half received an automated welcome DM with three suggested first actions, half received nothing. Both groups saw the same server.

What the data shows
— First-message rate within 48h: 19% (DM group) vs 12% (control).
— Members who completed at least one suggested action had a 2.1x higher 30-day retention than those who didn't.
— But overall 30-day retention difference between the two arms was a modest +3 points (17% vs 14%).

Why it happens
The DM gives a frictionless next step at the exact moment of peak intent — right after joining. The action-completion correlation is the headline, but note the direction of causality is murky: motivated people both complete actions and stick around.

The caveat
The +3 point overall lift is real but small, and DMs from servers increasingly trip spam instincts — one community reported rising 'DM disabled' rates that shrink the addressable group over time. Telegram has no native equivalent; bots can't cold-DM, which makes this tactic largely Discord-specific.

Open question: is the welcome DM actually changing behavior, or mostly identifying already-motivated joiners who'd have engaged regardless?