Experimentation Week is back this summer!
5 days. 5 launches.
At Statsig, we take pride in being the most advanced experimentation platform, and we’re committed to continuously innovating and pushing the boundaries.
To celebrate Experimentation Week ’25, we’re rolling out five new advanced capabilities.
Follow along on LinkedIn for daily drops throughout the week.
CURE: (variance) control using regression estimates is our advanced implementation of CUPED that lets you add arbitrary covariate data for regression adjustment.
Sequential probability ratio test: This is an advanced A/B testing technique that accelerates decision-making. It moves away from communicating p-values and instead computes ‘likelihood ratios’. Learn more in our docs.
This technique allows you to systematically test combinations of model hyperparameters to find the best-performing one.
You can treat experiment exposures just like any other event in metrics drilldowns and funnel charts, helping you debug faster and gain insights into the conversion paths of users exposed to an experiment variant.
Store prompts and models as configs and benchmark them against your eval dataset whenever you make a change. Then, ship to production as an experiment to measure the real user impact.