Measurement Nuances: Median vs. Average & Attribution

Why early data is noisy and how scores are computed robustly.

Outcome measurement is honest about uncertainty. Here's what's happening under the hood.

Median vs. average

Outcome scores use medians to reduce the influence of outlier days (a viral spike or an outage shouldn't define your result). Some on-screen displays show averages for readability — they answer different questions, so they can differ.

Early data is noisy

Right after a change, a few days of data can swing wildly. Wait for the full tracking window before drawing conclusions.

Attribution limits

A score is directional evidence, not proof. Seasonality, algorithm updates, and other changes you ship can all move traffic. When you want certainty, change one thing at a time.