How to A/B Test LinkedIn Outreach Messages

Banner with the title How to A/B Test LinkedIn Outreach Messages next to a simple chart illustration
Share

You wrote two versions of a LinkedIn message and you have a feeling one is better. Maybe the shorter note sounds friendlier, or the question at the end feels more natural. But a feeling is not proof, and every prospect who gets the weaker version is a chance you only get once. That is why it pays to A/B test LinkedIn messages instead of guessing.

An A/B test is simple. You send version A to one group and version B to a similar group, then compare the results.

The hard part on LinkedIn is volume. Most people send a small number of requests each day, so results come in slowly.

Small numbers also make it easy to crown a "winner" that only won by luck.

This guide shows you how to run fair tests with small volumes, what to test first, and how to read the results without fooling yourself.

Let's start with a quick overview.

A Quick Overview

Each part of your outreach moves a different number. Match the test to the number you want to improve.

What you test Where it shows up Number to watch
Connection note (or no note) Connection request Acceptance rate
First message after accepting Chat window Reply rate
Follow-up wording and timing Chat window Reply rate on follow-ups
Call to action Any message Positive replies or booked calls
Audience or list source Who you target Acceptance and reply rate

If you are not sure which numbers to track, our guide to LinkedIn outreach metrics and KPIs explains each one.

How to Set Up a Fair A/B Test on LinkedIn

A test is only useful if the two versions get a fair chance. Three rules keep it fair.

Change One Thing Only

If version B has a new opener, a new question and a new length, you will not know which change made the difference. Pick one variable. Keep everything else the same.

Split the Same Audience

Both versions should go to the same kind of people. Take one prospect list and split it in half, for example by alternating rows. Do not send version A to founders and version B to sales managers. That tests the audience, not the message.

Send in the Same Time Window

Send both versions during the same days. If version A goes out in a holiday week and version B goes out the week after, timing can change the result more than the wording does.

What to Test First in LinkedIn Outreach Messages

Start where the drop-off is biggest. For most people that is the connection request, because nothing else happens until someone accepts.

  1. Note vs no note. Some audiences accept plain requests more often. Others respond better to a short reason. Our guide on whether to add a LinkedIn connection request note covers the trade-offs.
  2. The opening line. Test a reference to the person's recent post against a reference to their role or company.
  3. Length. Test a two-line message against a four-line one.
  4. The question you ask. Test an easy question ("Is this on your list this quarter?") against a softer one ("Would a short example help?").
  5. Follow-up timing. Test a follow-up after a few days against one after a week. See how to follow up on LinkedIn for wording ideas.

Once acceptance looks healthy, move on to the first message after someone accepts. This guide on what to say after a LinkedIn connection is accepted gives you starting versions to compare.

How Many Sends Before You Judge?

This is where most LinkedIn tests go wrong. With small groups, random chance can easily make one version look better.

Here is a simple way to think about it. If version A got 6 replies from 40 sends and version B got 9 replies from 40 sends, that gap is only three people. A few people being busy that week could flip it.

Our suggestion, not a benchmark: do not pick a winner until each version has reached at least 100 people, and keep the test running for at least two full weeks. If the gap between versions is small, keep going longer or call it a tie.

For a more exact answer, use a free A/B test sample size calculator. Enter your current rate and the smallest change you care about. It tells you how many sends each version needs.

If that number is far more than you can send, test bigger changes. A completely different angle is easier to detect than one swapped word.

A Simple Tracking Sheet

You do not need special software. A spreadsheet with one row per prospect is enough.

Column What to record
Version A or B
Date sent The day the request or message went out
Accepted Yes or no, and the date
Replied Yes or no
Reply type Positive, neutral, not now, or no thanks

Add a summary row for each version that counts sends, acceptances and replies. Then divide to get each rate.

Track reply type, not just replies. A version that gets more "no thanks" answers is not a better message. If you want to see how acceptance compares across different approaches, our post on LinkedIn acceptance rate is a good next read.

How to Read the Results Without Fooling Yourself

  • Look at the gap and the group size together. A large gap on a small group is still weak evidence.
  • Do not stop early because you like the result. Decide the sample size before you start, then stick to it.
  • Check reply quality. More replies only help if more of them lead to real conversations.
  • Re-test winners later. Audiences change. A message that worked in spring may fade by autumn.
  • Keep a test log. Write down what you tested, the dates, the numbers and what you decided.

Sending fewer, better messages also makes each test cleaner. This post on why sending fewer outreach messages gets more replies explains why.

Quick Checklist

  • One variable changed between version A and version B
  • Same audience, split evenly
  • Same sending days for both versions
  • Sample size decided before the test starts
  • Tracking sheet with version, dates, acceptances, replies and reply type
  • Winner chosen only after enough sends, or the test called a tie
  • Result and decision written in a test log

For ready-made starting versions, browse these LinkedIn connection message templates and test them against your own.

Test Ideas, Not Templates: Linkyfy for Message Testing

When every message is written for one person, you are no longer testing word swaps. You are testing ideas, such as opening with a recent post versus opening with the person's role.

Linkyfy runs in the cloud and writes each LinkedIn message from the prospect's real profile and activity, not from a fixed template. You choose the angle and context, and every prospect still gets a message written for them. Sequences stop when someone replies, so a test never sends a follow-up to a person who already answered.

Built-in A/B testing is launching soon. Until then, you can run each angle as its own campaign and compare the results in your tracking sheet. See how LinkedIn message automation with Linkyfy works.

FAQs on A/B Testing LinkedIn Messages

What is an A/B test for LinkedIn messages?

It is a simple comparison. You send version A of a message to one group and version B to a similar group at the same time, then compare acceptance or reply rates to see which version works better.

What should I A/B test first on LinkedIn?

Start with the connection request, because nothing else happens until someone accepts. Test a note against no note, or two different opening lines, before you test follow-ups or calls to action.

How long should a LinkedIn A/B test run?

Our suggestion is at least two full weeks and at least 100 people per version. With smaller groups, random chance can make one version look better even when there is no real difference.

Can I test more than one change at a time?

You can, but you will not know which change caused the result. Change one thing per test. If your volume is low, test bigger differences, such as a completely new angle, rather than one swapped word.

Do I need special software to A/B test LinkedIn messages?

No. A spreadsheet with one row per prospect works well. Record the version, the date sent, whether they accepted, whether they replied and the type of reply, then compare the rates for each version.

What if both versions perform about the same?

Call it a tie and keep the version you prefer. A small gap on a small group is weak evidence. Use the result as a sign to test a bigger change next time.

Ravi Kotwani

Written by

Ravi Kotwani

CTO & Founder, Linkyfy.ai

Ravi Kotwani is the CTO and Founder of Linkyfy.ai. He has spent 16 years building web, mobile and AI products, and now helps founders and marketers grow with marketing automation, starting with LinkedIn outreach.

Follow on LinkedIn →