A/B testing: keep the message that works
2 versions of a step in one campaign. Keep the winner.
- +76% qualified calls
- No duplicate campaign
- LinkedIn and email
10,000+ GTM teams run their outreach on La Growth Machine
Use case€1.2Mqualified pipeline with multichannel outreach
InvoxUse case€430kpipeline from one LinkedIn + email campaign
Kodiak HubUse case$500kin pipeline generated with La Growth Machine
Use case30 to 35%of pipeline now from outbound, up from ~10%
Use case12%of replies converted
UtilaUse case15%+reply rate across campaignsWhat to test, and what to keep
Test your invitation note
Half get the invite with a note, half without.
Click A/B on your invitation step: version A without a note, version B with one, or 2 different notes. Half of your leads get each version, in the same campaign.
Find the opener that gets replies
Which first message works best after they connect?
On a LinkedIn message, click A/B and write a second version: another hook, another argument, another call to action. Change one thing and keep the rest, so you know what made the difference.
Test the subject line before you scale
“{{firstName}}, quick question” or “Idea for {{companyName}}”?
On an email step, test the subject line or the body. Run tests on several steps of the same sequence, LinkedIn and email alike: one audience, one set of stats.
Keep the winner for the leads left
B is performing better: the other 300 leads get B.
Channel Statistics shows the leads, replies and won leads of each version, plus opens and clicks for emails. Pause the version that loses, and new leads get the one that wins.
task_altA paused version stays in the step: resume it at any time.
qualified calls with an honest invitation note than with no note, in our A/B test on 9,000 leads (6.62% vs 3.77%).
Everything A/B testing can do
See everything A/B testing does expand_more
edit_noteWhat you can test
- LinkedIn invitation: 2 different notes, or a note against no note
- LinkedIn message: hook, argument, call to action
- Email: subject line and body
- Variables such as {{firstName}} in each version
call_splitHow leads are split
- 2 versions per step
- Half of your prospects get version A, half get version B
- Both versions live in the same campaign: no duplicate, no second audience
alt_routeWhere it works
- On several steps of the same sequence
- On sequences started from a template
- On a running campaign: pause it, add version B, resume
query_statsResults
- Each tested step opens into version A and version B in Channel Statistics
- Leads, replies and won leads for each version
- Opens and clicks for email versions
- The same results in Reports
pause_circleControl
- Pause a version from the editor, Channel Statistics or Reports
- New leads get the version still running
- Resume a paused version at any time
tips_and_updatesGood practice
- One change per test
- At least a week of data before you pick
- Test the step that most leads reach first
Run your first test today
2 clicks in the sequence editor, then read the results.
Open an invitation, LinkedIn message or email step click A/B write version B
Compare version A and version B pause the one that loses
Claude or ChatGPTLGM Analysis AgentAsk “Which message gets the most replies?” get the answer from your outreach data
Start from a template add a version B to any step launch
We've got you.
Your performance is our retention plan. Our team answers fast when you need help, and growing teams get their own Outreach Expert to review their campaigns and send a quarterly report with the actions that matter.
Talk to our teamStop guessing, start testing
Add a version B to the step your leads see first.
Frequently asked questions
Which plans include A/B testing? Can I try it for free? add
A/B testing is part of the Pro and Ultimate plans, and of the 14-day free trial. To test invitation notes, use LinkedIn Premium or Sales Navigator: a free LinkedIn account sends only a few invitations with a note each month.
What can I A/B test? add
LinkedIn invitations (2 different notes, or a note against no note), LinkedIn messages, and emails (subject line and body). Each step can have 2 versions.
How are prospects split between the versions? add
Half of your prospects get version A and half get version B, within the same campaign.
How do I pick the winner? add
Compare each version in Channel Statistics or Reports: leads, replies and won leads, plus opens and clicks for emails. Then pause the version that loses. New leads get the remaining version, and you can resume the other one at any time.
Can I add a test to a campaign that's already running? add
Yes. Pause the campaign, add a version B to the step, then resume: the test applies to the leads who haven't reached that step yet.
How long should a test run? add
At least a week, with one change at a time, so you know what made the difference.
Can I test several steps at once? add
Yes. You can run tests on several steps of the same sequence, LinkedIn and email alike. Change one thing per step.
What should I test first? add
Start with the step every lead sees, often the invitation, and judge each version on booked calls, not only on acceptance. In our 2025 test on 9,000 marketing leaders at large companies, acceptance barely moved (31%, 29% and 32%), but an honest invitation note brought 76% more qualified calls than no note: 6.62% of leads against 3.77%. The acceptance rate alone would have hidden the winner.
Why test inside one campaign instead of duplicating it? add
Both versions share one audience and one set of stats, so the comparison is fair and you don't split leads by hand. The test doesn't take a second active campaign slot, and a paused version can come back at any time.
Can I compare messages across all my campaigns? add
Yes, with LGM Analysis Agent, part of the Ultimate plan. Ask it from Claude or ChatGPT which message gets the most replies, and it answers from your own outreach data.