The verdict in three sentences
An A/B test compares two checkout versions to find which one truly converts better, with numbers to prove it. Run badly — too few visitors, several changes at once, stopped too early — it gives false conclusions that cost money. In 2026 the discipline is simple: one variable, a sufficient sample, two to four weeks, a significance threshold.
The clean method
The three mistakes that ruin a test: changing several things at once, stopping the moment one version "seems" to win, and testing on too few people.
| Rule | 2026 benchmark | Why |
|---|---|---|
| One variable per test | 1 button OR 1 step | Otherwise you can't tell what worked |
| Minimum sample | ~1,000 visitors/variant | Below this, chance dominates |
| Duration | 2 to 4 weeks | Covers cycles (payday, weekends) |
| Significance | ≥ 95 % confidence | Avoids false winners |
| Minimum conversions | ~200 per variant | Reliable basis to conclude |
The high-impact tests
Start with the changes most likely to move conversion. One test = one clear hypothesis = one metric to track.
| Test | Hypothesis | Tracked metric |
|---|---|---|
| "Pay with M-Pesa" vs "Submit" | A named MoMo reassures | Payment click rate |
| 1 step vs 3 steps | Fewer steps = less drop-off | Checkout completion rate |
| MoMo default vs neutral choice | The default guides the shopper | Share of completed payments |
| Fees early vs fees at the end | Transparency cuts leakage | Payment abandonment rate |
| Guest checkout vs account required | Zero friction = more sales | Overall conversion |
Mini case study
Need a professional website?
Kolonell builds websites that attract clients, optimized for the Sénégalese market. Free quote in 2 minutes.
Binta runs an online shoe store in Lagos with 4,000 checkout visitors/month. She tests "Pay with M-Pesa" (variant B) against "Submit" (variant A) over 3 weeks, 2,000 visitors per variant. A converts at 60 %, B at 65 % — significant at 96 %. She rolls out B. Across her 4,000 visitors, +5 points = +200 orders/month; at 24,000 FCFA (approx. 37 EUR) AOV, +4,800,000 FCFA/month, nearly 58 million FCFA per year.
FAQ
How many visitors do you need at minimum? About 1,000 per variant and 200 conversions per variant for a reliable conclusion. Below that, the result may be down to chance.
Can you test several things at once? No, not in a simple A/B. If you change button + steps + fees together, you'll never know which acted. One test, one variable.
How long should a test run? At least 2 to 4 weeks, to cover paydays and weekends. Stopping on day 3 because one version "wins" is the most common mistake.
What if the test isn't significant? Deploy nothing and keep the current version, or extend the test if traffic is low. An inconclusive result is a valid answer.
Do you need expensive tools? No. A well-built store lets you route traffic and track conversions; what matters is method rigor, not tool price.
Let's talk about your project. We set up clean A/B tests to push your checkout conversion up. WhatsApp +221 77 596 93 33.
Mohamed Bah
Fondateur, Kolonell
Passionate about digital and entrepreneurship in Africa, Mohamed has been helping Sénégalese businesses with their digital transformation since 2020. Founder of Kolonell, he believes every SME deserves a professional and accessible online présence.
