A/B testing

Test variants of invite notes, messages and InMails, read the results per variant, and pick a winner.

Invite notes, messages and InMails can have up to five versions: the step's own text (variant A) and up to four variants (B to E).

Set up a test

  1. Open the step in the sequence builder and choose Add variant. Traffic is split evenly; change the Share of each variant (1 to 99%). Variant A gets what the others leave.
  2. Write each variant. Variables and inline defaults work as usual. An InMail variant without a subject keeps variant A's subject.
  3. The step's fallback text covers every variant when a variable has no value for a lead.

Which lead gets which variant

The variant is chosen from the lead and the step, not at random on each send: a lead always gets the same variant of a step, even after a retry, a pause or a restart. Different steps choose independently, so a lead can get variant B of the invite and variant A of the follow-up.

Results

The campaign's Overview shows an A/B tests card for every step with variants:

ColumnMeaning
ShareThe configured percentage of leads.
SentInvites, messages or InMails sent with the variant.
AcceptedFor invites: leads who connected after receiving it.
RepliedFor messages and InMails: leads who replied after receiving it.
RateAcceptance rate for invites, reply rate otherwise.

Once every variant has at least 20 sends, GraphKit marks the one with the best rate as Leading. It does not pick a winner for you.

Pick a winner

Choose Pick as winner on a variant. Every lead that has not had the step yet gets that variant; leads already contacted keep what they received. This works on running campaigns. Resume split goes back to the shares you set.

HeyReach sends variants at random and does not document per-variant results or winner selection.

On this page