Campaigns
Different copy, images, send times, discount amounts — test them against each other. Edesy splits the audience, measures conversion, names the winner. No statistics PhD required.
Variants per test
Bayesian winner detection
Audience split
Equal allocation
Tests in drips
Optimize each step
Decision
On typical campaign sizes
Two versions of the same message — different headlines, body copy, CTA wording. Edesy randomizes which version each contact gets and tracks conversion per variant.
Same copy, different hero image or carousel card order. Find the visual that pulls clicks.
Variant A sends Tuesday 10 AM, Variant B sends Tuesday 6 PM. Same audience pool, same message — find when your audience converts best.
Test 5% vs 10% vs 15% discounts. Find the floor where extra discount stops adding incremental conversions.
Inside a drip, test individual steps independently. Steps 1, 2, 3 each get their own variant test. Iterate the weakest steps in isolation.
Edesy uses Bayesian statistics (not frequentist null-hypothesis testing) — more intuitive thresholds, better at small sample sizes typical of WhatsApp campaigns.
Set up a campaign, then click 'Add variant'. Duplicate the original and change ONE thing (copy, image, timing, CTA, discount). Edesy enforces single-variable testing to keep results interpretable.
When the campaign launches, recipients are randomly assigned to a variant. Default 50/50 (or even split across N variants). You can override to 70/30 for risk-averse tests.
Per-variant dashboard shows sends, opens, clicks, conversions. Bayesian 'probability of being best' updates live as data arrives.
Once a variant has >95% probability of being best, Edesy notifies you. Auto-promote mode shifts the remaining audience to the winner; manual mode waits for your decision.
Test results stored in your campaign history. Build up institutional knowledge over months: 'urgency CTAs convert 2.1x better for our audience', 'images with people outperform product-only by 35%'.
| Feature | Edesy | Typical platform |
|---|---|---|
| Built-in A/B testing | Add-on/premium | |
| Multivariate (2–4 variants) | 2 only | |
| Per-step testing in drips | ||
| Bayesian winner detection | Frequentist | |
| Auto-promote winner | Manual | |
| Test send time variants | ||
| Test discount amounts | ||
| Historical test archive | Manual notes |
Tested 10% vs 15% vs 20% cart recovery discounts. 15% won — beat 10% on conversion, matched 20% but at higher margin.
5% margin improvement at same volume, ~₹2L/mo recovered
'Book free demo' vs 'See it in action' vs 'Try a live class'. 'Try a live class' won decisively.
Demo booking rate up 47%
Tested 'Reply to talk to an agent' vs 'I can help right away — what's your question?'. Bot-first won.
60% reduction in agent escalations
Hotel exterior photos vs 'guest enjoying room' lifestyle photos. Lifestyle won by 38%.
Carousel CTR up from 12% to 17%
Tested 10 AM vs 6 PM vs 9 PM for loan reminders. 9 PM (after work) converted best for the audience.
Payment completion rate up 23%
Tested '7-day extension' vs '14-day extension' vs 'extend or 20% off' in churn-save flows.
'Extend or 20% off' converted 2x other options
A/B testing has been standard practice in email marketing for 15 years, but it's still rare in WhatsApp. Part of the reason is that most WhatsApp platforms don't support it natively — you have to manually create two campaigns, manually split the audience, manually compute the results in a spreadsheet. That's enough friction to discourage teams from testing systematically, so they ship one variant per campaign and hope for the best.
The cost of not testing is hard to see because you never see the variant you didn't ship. But the data is consistent: when teams start A/B testing WhatsApp campaigns, they typically find that one in three variants beats the original by 20%+ on conversion. Compounded across 50+ campaigns a year, the brand that tests every campaign ends up 2-3x ahead of the brand that ships one variant per campaign — without spending any more on infrastructure or audience.
Edesy's A/B testing is designed to remove the operational tax. You set up a campaign, click 'Add variant', change one thing, hit launch. Audience splits automatically; results show in real time; Bayesian winner detection tells you when you have a statistical decision; auto-promote mode shifts the rest of the audience to the winner without manual intervention. The whole loop is 5 minutes of work per campaign, and pays for itself many times over.
The Bayesian approach matters more than it sounds. Traditional frequentist A/B testing requires pre-committing to a sample size and waiting for 'statistical significance' (usually p<0.05). That's fine for huge web traffic but brittle for the smaller audience sizes typical of WhatsApp campaigns (often 5K-50K). Bayesian testing reports 'probability variant A is best' continuously — at 5K sends you might already have 92% confidence in a winner; at 20K you might have 99%. You can stop testing early when you're confident, or keep going when you're not. Much better fit for how WhatsApp campaigns actually work.
Free workspace, A/B testing built into every campaign. Most teams ship their first test within a day of signup.