To A/B test LinkedIn messages, split one prospect list at random into two equal groups, send each group a different version of one message, keep everything else the same, and compare acceptance rate and reply rate once each version has about 200 sends. LinkedIn has no A/B testing for the messages you send from your own account, so you run the split yourself or with an outreach tool that tracks reply rate per version.
Can you A/B test messages inside LinkedIn?
No. LinkedIn's built-in A/B testing lives in Campaign Manager and "compares the performance of two ad sets with separate budgets that differ by one variable." It is built for ads. It does not cover the connection notes, DMs, or InMails you send from your own profile or from Sales Navigator.
That leaves two ways to test outreach messages:
- By hand. Split your list in a spreadsheet, send version A to one half and version B to the other, and log every acceptance and reply against the version it came from.
- With an outreach tool. The tool splits the prospects who reach a step, sends each half a different version, and reports the result per version.
The method below works for both.
How to A/B test a LinkedIn message, step by step
- Pick one step. A connection note, the first message after someone accepts, or a follow-up. Test one step at a time.
- Change one thing. The opener, the call to action, the length, or the reason you give for reaching out. If you change two things and B wins, you won't know which change did it.
- Split one list at random. Both groups should come from the same list: same titles, industries, company sizes, and regions. Splitting by hand? Sort the list by a random number column and take alternate rows.
- Keep everything else the same. Same sender, same send window, same follow-up timing, same days of the week.
- Set your sample size before you start. See the table below. Deciding when to stop after you see the numbers is how most tests produce false winners.
- Measure the right number for the step.
- Connection note: acceptance rate (accepted ÷ requests sent).
- First message and follow-ups: reply rate (replied ÷ messages sent).
- For the final call: positive reply rate and meetings booked, because a message can get more replies and fewer meetings.
- Keep the winner and test the next thing. The winning version becomes your control, and your next test changes one new thing against it.
What to test on each step
| Step | What to change | Metric to compare |
|---|---|---|
| Connection note | With a note vs. no note; the reason you give for connecting | Acceptance rate |
| First message after accept | The opener (a question vs. a compliment vs. a direct ask); the call to action | Reply rate |
| Follow-up | Timing (2 days vs. 5 days); a new angle vs. a short bump | Reply rate |
| Whole sequence | The offer or the problem you lead with | Positive reply rate, meetings booked |
How many messages do you need for a LinkedIn A/B test?
About 200 sends per version is enough to spot a 10-point difference in reply rate. Smaller tests only catch big gaps. These are rough sample sizes per version (two-sided test, 95% confidence, 80% power):
| Version A reply rate | Version B reply rate | Sends needed per version |
|---|---|---|
| 20% | 40% | ~80 |
| 10% | 20% | ~200 |
| 20% | 30% | ~290 |
| 10% | 15% | ~680 |
What this means in practice:
- ~100 per version catches only large differences. It is enough to drop a message that clearly doesn't work, not to pick between two decent ones.
- ~200–300 per version is where most useful LinkedIn tests land.
- Small differences (a 3–5 point lift) take more sends than most LinkedIn accounts can make in a month. Test bigger changes instead.
A worked example: version A gets 33 replies from 96 sends (34%) and version B gets 21 from 96 (22%). That looks like a clear win, but at 96 sends each it falls just short of significance at 95% (p ≈ 0.055). Keep sending both until each passes about 250, then decide.
How to test which lead list works best
To test lead lists, send the same messages to random, equal-sized samples from each list, then compare positive reply rate and meetings booked per list. It is the same test with the roles swapped: hold the message constant to test lists, and hold the list constant to test messages.
- Same messages, same sender, same dates. If the copy differs between lists, you can't tell whether the list or the copy made the difference.
- Equal random samples. Take the same number of prospects from each list, at least 200–300 each.
- Record which list every prospect came from. In a spreadsheet or CRM this is a "source" column. In an outreach tool that can compare prospects by where you found them, the results per list come out on their own.
- Give it about two weeks. Follow-ups need time to land before you count replies.
- Compare the full funnel, not just replies. On LinkedIn: acceptance rate, reply rate, positive reply rate, meetings booked. On email, also check bounce rate; a list that bounces more than 3–5% of emails has a data problem.
- Break ties on meetings or revenue. A list that gets more replies but fewer meetings is the worse list.
Here's how a list test can look once the numbers are in:
| List A (conference attendees) | List B (Sales Navigator search) | |
|---|---|---|
| Contacted | 250 | 250 |
| Accepted | 120 (48%) | 150 (60%) |
| Replied | 30 (25% of accepted) | 33 (22% of accepted) |
| Positive replies | 14 | 9 |
| Meetings booked | 6 | 3 |
List B got more acceptances and more replies, but List A booked twice the meetings. List A wins.
Already sent hundreds of messages? Compare them instead
If you have months of outreach behind you, you don't need a new test to learn what works. Sort the messages you've already sent by type and compare the reply rate for each type.
- Pick one angle to compare. Opening line, the ask, tone, or how personal the message is. Or compare who you sent to: job title, industry, company size, or where you found them.
- Describe 3–5 types in a sentence each. For opening lines: "asks a question about their process," "compliments something they posted," "asks for a call straight away." Every message should fit exactly one type.
- Sort every sent message into a type. In a spreadsheet, that's a column you fill in by hand. An outreach tool can read the messages and sort them for you.
- Compare reply and meeting rate per type. Look at first messages and follow-ups separately, because a follow-up gets replies for different reasons than an opener.
- Check the messages behind the winner. Read a handful before you believe the number. If a type only has 15 messages in it, treat the result as a hint and confirm it with an A/B test.
Use each method for what it's good at:
| Compare past messages | A/B test | |
|---|---|---|
| Question it answers | "What has worked so far?" | "Is this new version better?" |
| Needs new sends? | No, uses your history | Yes, ~200 per version |
| Controls for other differences? | No, so types may differ in who received them | Yes, one random list, one change |
| Best for | Finding patterns and ideas to test | Confirming a change before you roll it out |
A good loop: compare your past messages to find a pattern, then A/B test it to confirm.
Which tools A/B test LinkedIn messages?
Most A/B testing in outreach tools is built for email. Here's what each option does for LinkedIn messages:
| Option | LinkedIn A/B test | Email A/B test | How the split works | Results per version |
|---|---|---|---|---|
| Sliq | Yes | Yes | A "Send A/B" step splits prospects evenly between two versions | Reply rate and sends per version; promote the winner to everyone |
| HeyReach | Yes | Via integrations | "Add Message Variation" inside one LinkedIn campaign | Acceptance rate and reply rate per variation |
| Instantly | No | Yes, up to 26 variants per step | Variants distributed evenly across the campaign | Reply, click, or open rate; can switch off weaker variants automatically |
| Spreadsheet | Yes, by hand | Yes, by hand | You split the list and send each half yourself | Whatever you log |
How to A/B test and compare messages in Sliq
Sliq runs LinkedIn and email outreach as a flow of steps: connection request, first message, follow-ups, and emails. Any message step can be a Send A/B step:
- Two versions, split evenly. Sliq sends version A to half the prospects who reach the step and version B to the other half.
- Reply rate per version. Each version shows its reply rate and how many it has sent, on the flow and in the campaign summary.
- Promote the winner. When one version has enough sends and a clear lead, promote it and everyone who reaches that step from then on gets it.
- The full funnel per campaign. Contacted, accepted, replied, interested, and meetings booked.
For what you've already sent, pick what you want to compare, like opening line, ask, job title, industry, or where you found the prospect. Sliq sorts your sent messages and prospects by type and shows the acceptance, reply, and meeting rate for each, so a lead-list test is one comparison by where you found them.
See how it works on the outreach analytics page, or try Sliq free.
FAQ
Can you A/B test LinkedIn messages?
Yes, but not inside LinkedIn itself. LinkedIn's built-in A/B testing lives in Campaign Manager and compares ad sets, so it does not cover the connection notes, DMs, or InMails you send from your own account. To A/B test those, split one prospect list at random into two groups, send each group a different version of one message, and compare acceptance rate and reply rate. You can do this by hand with a spreadsheet, or with an outreach tool that splits the list and tracks reply rate per version, such as Sliq.
How many messages do you need to send for a LinkedIn A/B test?
Plan on about 200 sends per version to detect a 10-point difference in reply rate (for example 10% vs 20%) with reasonable confidence, and about 300 per version to detect 20% vs 30%. Around 100 per version is only enough to catch a large gap, such as 20% vs 40%. Decide the number before you start and do not call the test early.
What should you A/B test first on LinkedIn?
Test the step with the most volume first. That is usually the connection note (measured by acceptance rate), then the first message after someone accepts (measured by reply rate), then the first follow-up. Within a step, test the opener or the call to action before smaller edits like length or tone.
How do you test which lead list works best?
Send the exact same messages, from the same sender, over the same dates, to random samples of equal size from each list, then compare positive reply rate and meetings booked per list. Use at least 200 to 300 prospects per list and give it about two weeks so follow-ups finish. Hold the message constant when you test lists, and hold the list constant when you test messages.
Which tools A/B test LinkedIn messages?
Sliq splits a LinkedIn or email message step into two versions, sends each to half the prospects who reach that step, and shows the reply rate for each version so you can promote the winner. HeyReach offers message variations inside a LinkedIn campaign. Email-first tools such as Instantly A/B test email steps but not LinkedIn messages.
Can you find out which messages worked from ones you've already sent?
Yes. Pick one angle, such as your opening line, describe three to five types in a sentence each, sort every sent message into a type, and compare the reply and meeting rate for each type. It uses your history, so it needs no new sends, but types can differ in who received them, so confirm a promising pattern with an A/B test.