What You'll Learn
CUPED is an incredibly powerful technique to speed up your tests, if you have pre-experiment data on returning visitors.
In this talk, you will:
Visualize the relationship between power, variance, and sample size.Understand how CUPED reduces the variance in your target metric.Observe the simulational proof of CUPED reducing variance in data.See what you need to apply CUPED in your own A/B tests.This Webinar Is For You If
This talk is for you if:
You have been hearing about CUPED but never fully understood what it is about.You are dealing with metrics like revenue that have a high variance, leading to large sample sizes.More often than not, your tests are underpowered and do not reach significance.You are curious about understanding advanced statistical techniques through interesting analogies.After this talk, you will have the relevant intuition to decide if you can use CUPED for boosting power in your tests or not.
Hi, I am Ishan Goel, a statistician, educator and founder of Thinking Bell.
I previously led data science research at VWO, where I built VWO's Stats Engine.
Today, I teach experimentation statistics through analogies, simulations, games, doodles and real-world problems, making difficult statistical ideas intuitive and useful.
I have also taught experimentation and statistics to cross-functional teams across product, marketing and data, including a 60-person training programme at The Economist.
My goal is simple: help people reason confidently about experiments without turning them into statisticians.
Stat-Talk #9