What's a good win rate for a CRO program with AB Testing?

At ConversionBoost in Copenhagen, Lucia recently gave a talk—and the most asked question afterward was: “What’s a healthy A/B test win-rate?”

Let’s break it down.

First, What Is Win-Rate in A/B Testing?
Your win-rate is the percentage of A/B tests that result in statistically significant improvements.

For example, if you run 100 experiments in a quarter and 20 deliver significant results, your win-rate is 20%.

So, when we say "win," we mean a variation that beats the control with statistical confidence.

Why keep track of win-rates in experimentation programs?

While the goal of experimentation isn’t just to chase wins, tracking win-rate can reveal a lot about the health of your CRO program.

A few things your win-rate can indicate:

• Research quality – A high-quality hypothesis informed by good data usually leads to better outcomes.
• Program maturity – Win-rates tend to gradually decrease as programs evolve and easy wins get hard to find.
• MDE alignment – If your tests frequently show no significant difference, it might mean your Minimum Detectable Effect (MDE) is too high for the changes you are making. (We help clients calculate and adjust for this.)

A sudden drop in win-rate?  Worth investigating.

What's an "Immature" Win-Rate?

When experimentation is new at a compan y—especially with expert support— win-rates often start off high, sometimes exceeding 50%.

Why? Because you're likely testing obvious, high-impact changes. This “low-hanging fruit” phase often results in fast, easy wins.

As your program matures and the easy wins are already captured, your win-rate will naturally decline. 

What’s a Healthy Win-Rate?

At our agency, we typically aim for a 30% win-rate across clients.

But really anything within 10%-50% can be okay. Depending on the Experimentation Maturity, assuming a high quality of research and data coming in to form hypothesis and A/B tests on. 

Should I focus on maintaining a high win-rate?
NO! When you focus on wins only, you'll avoid explorative, innovative ideas because they bring risk. When you focus on wins you'll only go into optimization directions that you are sure will work for you. So sometimes it can be very helpful to also measure learning rate: the % of experiments that lost and you learned from them. Those two together make for a more holistic approach. 

Win-rates of big companies
(the sources are quite outdated, but still interesting to refer to)

(Kohavi, Crook and Longbotham 2009) (Kaushik 2006) (Kohavi, Deng and Longbotham, et al. 2014) (Manzi 2012, Thomke 2020, Moran 2007)