How to Use Dripify to A/B Test a LinkedIn Follow-Up Message

The short answer: if people accept your connection requests and then ignore the first message, test that message, not the invitation: Dripify splits new leads at random across up to 3 versions of a Message step, on the Advanced plan at $99 a seat a month or $79 billed annually (A/B testing article, pricing page, checked on 26 October 2026). The steps: load enough leads, build a sequence that waits for acceptance, turn the first message into variants that differ in one thing, launch, read each variant's replies in step analytics, switch off the losers, and confirm the winner against meetings in your CRM.

Disclosure: I may earn a commission through some links on this site. It does not change what this page says.

Before you split anything:

1. The sample. For example, 300 people accept and reach the message step; three variants give each about 100, so a single reply moves a variant's rate by a full percentage point, and a gap of two or three replies is noise. Dripify's own A/B testing article warns that small samples mislead.
2. The message you already send. Keep it, word for word, as variant A. Last month's reply rate on it is your baseline, and the new variants have to beat that, not each other.
3. What counts as a win. Decide now whether you are judging replies or booked calls. Dripify counts replies; your CRM counts calls.

The plans behind this are in the Dripify glossary entry.

What you need

• Advanced, or the 7-day trial, which runs on Advanced features; A/B testing is a beta for Advanced users, per the A/B testing article.
• A LinkedIn account connected to Dripify. A free account is fine for testing messages; testing the invitation note needs LinkedIn Premium, because free accounts get only 5 personalised notes a month.
• A list of a few hundred people, from a LinkedIn or Sales Navigator search, an event, or a CSV of profile URLs.

Splitting a message into variants sends no extra messages; the same leads are divided between versions. The account risk stays what it was, and LinkedIn is clear about it: its prohibited software page says it keeps improving its technical defences against automation tools, and section 8.2 of its User Agreement names bots that send messages.

Step 1: Load the leads

Paste a LinkedIn search URL and choose how many people to add, or upload a CSV that contains LinkedIn profile URLs, as the lead import article describes. A free LinkedIn or Recruiter search stops at 1,000 profiles; Sales Navigator allows 2,475.

To top the list up later without retyping it, connect the Dripify MCP server to Claude: generate an API key under Settings, Integrations, then add https://app.dripify.com/api/v1/mcp with that key as a custom connector, per the MCP article. Give the assistant a CRM export of target accounts and ask it to assemble up to 1,000 LinkedIn profile URLs as a new lead list for this campaign. Before it uploads, a person reads the list and strikes current customers and open deals; the upload is the only write action the MCP offers, so that review is the whole human check.

Step 2: Build a sequence that waits for acceptance

Create the sequence from a template or from scratch: view profile, send invite, then a condition on acceptance with a delay before the message. Dripify cannot delay the first action, so its sequence article suggests opening with View profile and putting the delay after it. Close every branch, or the campaign will not launch.

Step 3: Turn the first message into variants

Find the Message action that follows acceptance and click Run A/B test. That creates variant B; the button then reads Add variant, and a third click gives you C. Variants are lettered automatically and cannot be renamed, and a sequence with a test running cannot be saved as a template, per the A/B testing article.

Change one thing per variant: the opening line, the question at the end, or the length. If B changes the opening and the question at once, a win tells you nothing you can reuse. Dripify's Improve with AI button, active once a draft reaches 25 characters, rewrites in a chosen style (Professional, Friendly, Convincing, Curious or Helpful) and shows the original beside the new version, per the AI writing article. Use it to produce B from A, then read both aloud.

Step 4: Name it, add the reply webhook, launch

Name the campaign after the test ("Follow-up test: question vs case"). If your CRM should hear about replies, add the webhook trigger "After LinkedIn reply is received", which sends the whole conversation, then click Save & Launch, as the launch article describes.

Keep the daily limits sensible while the test runs. Dripify's limits article suggests 20 to 30 connection requests a day for established accounts and 10 to 15 for new ones, rising 10 to 20% a week after the first one or two.

Step 5: Read each variant, then wait

Each variant shows its own sent, replies and failed counts as tags on the action, and step analytics add acceptance and reply rates for each step once the campaign is live. Step analytics are a beta on Advanced and appear only on campaigns created after the feature's release, per the step analytics article.

Do not edit a running variant's wording mid-test; a change resets what you are comparing. Wait until each variant has reached the sample you set before step 1.

Step 6: Keep the winner

Toggle the weaker variants off and save. A switched-off variant stops receiving new leads and keeps its data, and a deleted one still counts in the campaign's totals, per the A/B testing article. At least one variant has to stay on.

Where Dripify's docs draw the lines

• Free LinkedIn accounts cannot test invitation notes. The limit of 5 personalised notes a month makes a note test impossible.
• Tests block templates. Save the sequence as a template before you add variants, not after.
• Betas move. Dripify says step analytics may change plan on full release.

Check the winner against meetings

The free check needs only your CRM. Before launch, note how many meetings last month's message produced per 100 accepted connections; after the test, count the same for the winning variant over its first 100. If it won on replies and lost on meetings, it was a more curious reply, not a better message.

Dripify picks the winner on replies that Dripify counts, and that counting, step analytics plus the A/B results, is what the extra $20 a seat on monthly Advanced pays for; a reply that never becomes a meeting scores the same as one that does. Later, if you want that comparison run continuously against revenue rather than replies, a causal attribution tool such as Causality Engine does it (disclosure: I founded it).

FAQ

Can I A/B test the connection request note?

Only with LinkedIn Premium. On free accounts Dripify disables the test for the Send Invite action, per the A/B testing article.

How many variants can run at once?

Three per action, lettered A, B and C. Dripify caps it there because more variants mean fewer leads in each.

Does switching a variant off delete its results?

No. It stops new assignments and keeps the data; even a deleted variant stays in the campaign's overall analytics.