How do I A/B test LinkedIn messages?
Pick one variable: the first line, the ask, blank versus note, or wait time after accept. Leave everything else still. If you change the segment and the copy, you will attribute the wrong win.
Split a similar list, not "CEOs" versus "interns." Send enough that one lucky thread cannot fool you. For a small team, 40 to 80 first messages per variant is a fair look. For invites, watch accepts over a week, not overnight.
Write the variants before you start. Do not keep editing the "losing" one mid-test. That is not a test.
On LinkedIn you cannot always know who saw the note. Use accepts and replies, not opens. For email, keep the tests in the email tool. Do not mix denominators.
When a variant wins, freeze it as the new default and test something else. Permanent "testing" is how nobody on the team knows what you send.
Frequently asked questions
You can, and you will wait longer for a clean read. Two is enough until you have volume.
Only after copy and targeting are stable. Time-of-day effects are smaller than "wrong person."
Stop. Fix the list. Testing two bad pitches is a waste of the weekly cap.
No. You are not a lab. You are trying to be clearer.
Run the outreach from your own LinkedIn account
Omentir finds ICP-fit buyers, drafts connection notes and messages, and keeps replies in one inbox. You still choose the daily send limits.
Try Omentir