A/B testing WhatsApp messages: how to learn which message sells more, with two campaigns and one element
A/B testing on WhatsApp with the campaign tools you already have: two templates to two halves of the list at the same time, one element changed, and how to read replies and orders after 48 hours — with five elements worth testing and three mistakes that void the result.
On this page 4
The short answer
With two campaigns, not one: template A to half the list and template B to the other half, at the same hour on the same day, changing one element only (the first sentence, the presence of the name, or the button). Then read each campaign's replies from the unified inbox after 48 hours, and its orders by the campaign tag. The rule: one element, at least two hundred per half, and no reading the result before two days.
Also asked as: «اختبار رسالتين واتساب» · «اي رسالة تبيع اكتر» · «كيف أعرف الرسالة الأفضل للحملة» · «A/B test واتساب»
Example: only the first sentence changes
{{1}}, end-of-season offers have started: up to {{2}} off until {{3}}. Order here: {{4}}.
{{1}}, the last {{2}} pieces of the season, up to {{3}} off. Order before they run out: {{4}}.
The name, the discount, the link and the button are the same in both. What changes is the first sentence: a general offer versus scarcity. Any difference in replies is down to it alone.
The test in five steps
- Split the list in two with two labels
From the subscriber manager: labels "test-A" and "test-B" on those who opted in, alternating rather than in date order — otherwise you test old customers against new ones, not message against message.
- Get both templates approved a week ahead
Two templates at Meta differing in one element. Approval is free and takes minutes to two days.
- Two campaigns at the same hour
Campaign one: template A → label "test-A". Campaign two: template B → "test-B". Same hour, same day, because the hour alone changes replies.
- Tag the replies for 48 hours
Everyone who replies is tagged "reply-A" or "reply-B" in the unified inbox according to their label, and whoever orders is tagged "order-A" or "order-B". The AI agent answers everyone from your files; tagging is the team's.
- Read three numbers per half
From the campaign report: delivered and read. From the tag filter: replies and orders. The winner is whichever brought more orders, not more replies — a provocative message brings angry replies too.
Five elements worth testing
| Element | A versus B | What it measures |
|---|---|---|
| The first sentence | General offer / scarcity / a question | Replies |
| The name | "{{1}}," / no name | Replies and reports |
| Button versus link | "Order" button / a link in the text | Orders |
| Timing | Noon / after the evening prayer — same message | Read and replies |
| The image | A product image / text only | Orders — at the same cost |
Two elements at once: changing the sentence and the button in one test gives a winner you cannot explain. A different day per half: Thursday replies more than Monday with the same message. Reading after two hours: half the replies come on day two — wait 48 hours.
Frequently asked questions
How many customers do I need for the test to mean anything?
At least two hundred per half. Below that, a difference of five replies is chance, not a result. If your list is smaller, test across two consecutive campaigns two weeks apart instead of two halves on one day — same element, same time of week.
What do I read from the campaign report and what from the inbox?
The campaign report gives each campaign its sent, delivered, read and failed counts. Read measures the timing and the sender name, not the text, because the customer opens before reading. Reply and order are what measure the text, and they are counted in the unified inbox with a tag per half: whoever replied to A is tagged A. There is no reply counter in the campaign report itself.
Does the test cost more than one campaign?
No. The same list is split in two, and each customer gets one template at Meta's rate for their country on your account. What is added is the approval of a second template at Meta — which is free.
Which element do I test first?
The first sentence. It is what shows in the notification before the chat is opened, and it moves replies most. Then: the presence of the name, then a button versus a link, then send time, then an image. One element per test, or you will not know which one worked.
What is a good reply rate?
We publish no benchmark because we have not measured campaign replies across enough businesses to stand behind a figure. The benchmark is your other half: the test says which message is better, not whether your messages are good by the market's measure.
Read next
Try the agent before you pay anything
The agent that will answer your customers is the one answering you now: ask it from the chat button here or on WhatsApp. Your account is free with no credit card, and the free trial on your own number or page is activated after you contact our team on customer-service WhatsApp.