The Performance Page – Positive and Negative Experiments

Performance answers a different question from Health. Health asks what is *wrong*; Performance asks how the running experiments are actually *doing*. The two sit together in the menu and share the same scope tabs, because they are two ways of watching the same programme.

Two tables

Running experiments are split into Positive Experiments — those with a positive impact so far — and Negative Experiments, those currently producing a worse outcome than the control.

Either table can be expanded to fill the view when you want to work through it, and you can switch straight to the other while expanded. Rows expand individually for more detail without leaving the page.

Why the negative table matters most

The negative table is the one to look at first, and it is the reason this page exists.

A winning experiment costs you nothing while you deliberate — it is making money in the meantime. A losing experiment is actively sending a share of your traffic to a worse experience, and every day it runs has a price. Most optimisation programmes are much quicker to notice winners than losers, which is exactly the wrong way round.

That said, do not stop a test the moment it dips. Early data is noisy and a negative trend often reverts. Check the experiment’s Statistics tab before acting: look at whether the result is significant, at the confidence interval, and at whether the power analysis says the experiment can resolve the difference at all.

How it differs from the Experiment Library

The library is a complete inventory, filtered by status. Performance is a judgement about direction, and only covers experiments currently running. Use the library to find a specific experiment; use Performance to see which way the running ones are going.

Notes

• An experiment appearing in either table is a trend, not a verdict. Significance lives on the Statistics tab.
• A trend on an experiment whose traffic split is flagged as faulty is meaningless. Check Health first.
• Scope tabs work the same way as on Health: everything you can reach, or only your own experiments.