A/B testing campaigns

🎯 Goal

Find out which message gets more reads and replies, using real numbers instead of guesses.

🧰 What you need

• Two approved templates with the same purpose (for example two offer wordings), or one template you fill two ways
• An audience of at least a few hundred for meaningful results

How it works

The audience is shuffled and split when the campaign sends. Version A goes to one part, version B to the rest (you choose the share for B, 10–90%, in steps of 5). Each recipient gets exactly one version. Version B can be another template, or the same template with different values (another image, another offer value). The report shows Sent, Read and Replied per version and marks the leading one by reply rate.

Steps

1. Create two templates, for example diwali_offer_a (“20% off all sweet boxes”) and diwali_offer_b (“Free delivery on all Diwali orders”). Get both approved.
2. In the campaign wizard’s Message step, choose template A as usual and map its variables.
3. Turn on the A/B test card below the message. Under Version B, pick template B (or the same template) and map its variables.
4. Drag the slider: 50% get version A · 50% get version B.
5. Review and send. After a day or two, open the campaign’s View report. The A/B test table shows versions A and B, with the better reply rate marked leading.

login.meapi.in/campaigns/1/report

• The campaign report: status, progress, rates and messages over time.
• For an A/B campaign, the A/B test card lists versions A and B with Sent, Read and Replied.
• The version with the higher reply rate carries a leading badge.

Test one thing at a time

Good tests change one element:

• the first line (it shows in the notification),
• the offer (discount vs free delivery),
• the button (Shop now vs See offers),
• an image vs no image.

Small test, then roll out

Run the test with 20% B on a smaller segment first. Then send the leading version to everyone else, using Retarget or a new campaign that excludes the test group’s tag.

Common mistakes

• Tiny audiences. With 50 people per version, one reply swings the result. Aim for hundreds.
• Judging too early. Wait at least 24 hours. People read WhatsApp through the day.
• Different categories (Marketing vs Utility) for A and B. The costs and delivery behaviour differ, so the test isn’t fair.

What’s next

Analytics and reports →Read every number.