Reliability — When Some Units Never Fail

A life test almost always ends before every unit has failed. Those survivors are censored: you know they lasted at least that long, and nothing more. Counting them as failures makes the product look worse than it is; deleting them makes it look better. Both are silent — the curve still gets drawn.

The Weibull shape β names the failure mode: below 1 means early-life failures (assembly, infant mortality), around 1 means random external causes, above 1 means wear-out. B10 — the age by which one unit in ten has failed — is what warranty and maintenance intervals are usually built on.

The screen also shows what you would have got by ignoring censoring. Watch which row moves: B10 barely changes, while median life, mean life and β are all wrong. That is exactly why this mistake survives review.
Try it in the app
Try: Reliability tab → Load example → make sure the status column is selected → Analyse. Then set the status column to "none" and compare B50 and β.
Watch this on our YouTube channel

Run this on your own data

Free tier, no signup needed. Works offline in the browser.

Open the app