Measured and modelled: why we keep them apart.
When a circuit breaches, it is tempting to put a big number on it. That number is usually a guess wearing a suit.
There are two kinds of number in any conversation about a network failure. The first is what happened. The circuit was below its SLA for a measured period. The breach is time-stamped. The credit is calculated by rule from the contract. Run it again tomorrow and you get the same answer. This is measured, deterministic, reproducible and defensible. The second is what it might have cost. This is modelled. It is built from assumptions, and reasonable people can argue about every one of them. It is useful for context. It is not fact. Parallax keeps these two apart, on purpose. Measured fact goes on the record: performance against the SLA, the breach, the credit due. Any modelled estimate stays clearly labelled as an estimate, and never touches the measured record. The discipline matters because the moment a measured report borrows a number from a model, the whole thing becomes arguable. A credit you can defend line by line is worth more than an impact figure you cannot. Certainty is the product.
We report what happened. We are glad to help you think about what it meant. We never confuse the two.