# How Much Traffic Do I Need To Start Doing A/B Testing?

I’d like to start measuring improvements quantitatively, but I don’t think I have enough traffic. What’s the point when I can start doing real A/B Testing?

Improving a site’s effectiveness is a guessing game, that is, *until you have enough traffic* to demonstrate a difference between two versions. Version A, version B, that’s where A/B Testing (or split-testing), comes from.

When do you know you have enough traffic to make that measurement count?

You need a lot of traffic, but if you think you can make a big difference, you don't need as much.

## If Version B Makes a Bigger Difference, You Need Fewer Visits

Optimizely, a popular A/B Testing tool, comes up with these numbers in their online sample size calculator. *(Note: Optimizely bases their numbers on data from the A/B tests they run for everyone, and so other online sample size calculators will give you different numbers.)*

**You’ll need 26,000 visits**, or 13,000 for each version, if you’re currently converting 3% of your visitors and you think you can increase that **by 20%**. See the calculation on Optimizely with a 20% increase.

**But you’ll only need 10,600 visits**, or 5,300 for each version, if you’re currently converting 3% of your visitors and you think you can increase that **by 30%** instead. See the calculation on Optimizely with a 30% increase.

That difference comes from the fact that a bigger difference between two versions will take a smaller sample size to measure.

**You might ask:**

How can I reach a 30% conversion rate (or more) without trying all the nasty tricks that people pull on visitors for them to convert.

Well, it requires that you think deeply about **helping the visitor make some progress**.

*Be on the lookout in the next few weeks for articles on how to do that, and a breakdown of how to to calculate whether a Version B will be better.*

Stay Sharp!

—

Pascal Laliberté

@pascallaliberte