Testing on WhatsApp is constrained in ways email is not: each variant is a separately approved template, lists are usually smaller, and every test send costs money and audience goodwill. That argues for fewer, better-designed tests.
Test one thing
Two variants differing in offer, tone, length and call to action tell you nothing about which change mattered. Isolate a single variable, even though it feels slow. On a smaller list, a muddled test is worse than no test, because it produces a confident wrong conclusion.
What is worth testing, in order
- The offer itself. Far larger effects than any wording change. Free delivery against ten percent off will move numbers that a rewritten opening line never will.
- The call to action. Reply-based versus link-based, and which single action you ask for.
- Send time. Cheap to test and often surprisingly consequential.
- Message length. Short and direct against context-then-ask.
- Personalisation depth. Name only, against name plus last purchase.
Emoji, punctuation and small phrasing tweaks are last. They produce effects too small to detect at the list sizes most businesses have.
Measure replies and conversions, not reads
Read rates on WhatsApp are high and similar across variants, so they discriminate poorly. Replies and completed actions are where real differences show.
Sizing the test
If your whole list is 2,000 people, a 50/50 split of a 400-person test gives 200 per arm. At a 4% reply rate that is eight replies against, say, twelve — a difference well inside noise. Be honest about this: below a few hundred responses per arm, most differences you observe are random.
Two practical responses. Test bigger effects, where a real difference is large enough to see at your size. And run the same test across several campaigns, accumulating results rather than deciding from one send.
Practical mechanics
- Get both templates approved before you plan the send date.
- Randomise assignment. Splitting a list by upload order correlates with signup date, which correlates with engagement.
- Send both arms at the same time, or you have tested timing as well.
- Hold back the remainder of the list until you have a result, then send it the winner.
- Write down the hypothesis and the decision rule before sending. It prevents reading whatever you hoped for into the numbers.
Know when to stop testing
If three consecutive tests come back inconclusive, the variable is not where your gains are. Move up the list — usually to the offer — rather than continuing to refine copy. And once something wins clearly, adopt it and leave it alone; endlessly re-testing a settled question spends audience for no information.