Measure instead of meeting: provided you have enough visitors.
An A/B test shows two versions of a page to different visitors and measures which one performs better. This way you replace an opinion with a measurement. The condition is sufficient traffic: with few visitors it takes months before a difference is reliable, and until then every outcome is a coincidence.
IN COMMON LANGUAGE
The condition that is often skipped
To demonstrate a difference of a few percent, you need hundreds of conversions per variant. With ten applications per month, this will take years.
That makes A/B testing unsuitable for most SME websites. That is not a shortcoming but arithmetic, and it is better to know this in advance than to take a coincidental result seriously afterwards.
WHY IT MATTERS
What you do instead
If there is little traffic, you can make further progress in removing objections that you already hear in conversations. That is not a measurement, but it is more targeted than gambling.
And you can do qualitative testing: ask three people who don't know your site to say out loud what they do. That takes an afternoon and yields more than a test that is not completed.
- Enough conversions per variant — otherwise the result is coincidental.
- One change at a time — otherwise you don't know what it did.
- Determine in advance how long — stopping as soon as it suits you is self-deception.
- With little traffic: remove objections — more targeted than testing.
- With little traffic: have it tested out loud — three people, one afternoon.
FREQUENTLY ASKED QUESTIONS
More about an A/B test
How many visitors do I need?
Rule of thumb: several hundred conversions per variant to reliably demonstrate a small difference. Translate that into your own numbers before you start; The conclusion is often that testing is pointless now.
Do you do A/B testing?
Only for customers with enough traffic. Below that, we advise against it and prefer to look at what is missing in terms of content: that demonstrably yields more with those volumes.
Can I test button colors?
Possible, but that is rarely where the profit lies. Missing prizes, missing evidence and overly long forms carry far more weight than a colour.
CONTINUE READING
Of course, continue reading about this subject.
Test whether improve?
Tell us how many requests you receive per month. We say whether testing makes sense or whether there is something better.