Prediction ledger

Any vendor can claim their product predicts outages. Almost none will tell you how often they were right on your infrastructure, because they do not measure it.

The ledger writes down each warning the risk engine raises, then goes back and checks. Did an incident follow? How much warning was there? A prediction that led to nothing is recorded as a miss, sitting in the same table as the calls that landed.

What you get is two numbers you can audit: recall, the share of real incidents called in advance, and median lead time, the typical warning you received. Both live in your dashboard and nowhere else — this is your operational data, not a public scoreboard. If they get worse, you see it before we do.