Drivers¶

A driver is an input metric that explains part of why an anomalous metric moved.

FinSentry works out drivers by looking at the anomalous metric's formula and testing its inputs one at a time.

How it works¶

For an anomalous metric:

1. FinSentry follows its formula down to the underlying input metrics. If an input is itself calculated, it follows that formula too, until it reaches the raw metrics at the bottom.
2. For each of those inputs, it asks: if this input had stayed at its recent average, and everything else stayed as it actually was, how much of the move would disappear? That amount is the input's contribution.
3. Inputs with no effect are dropped. The rest are ranked by size, and each gets a share of the total (the shares add up to 100%).

Each driver is stored with its recent average and its current value, so you can see how much it moved.

Because inputs are tested one at a time, contributions are a guide rather than an exact split when inputs interact (for example inside a ratio). The shares are the reliable "which input mattered most" signal.

Multi-month anomalies¶

When an anomaly spans several months, each driver's contribution is added up across those months and the shares are recalculated.

Where drivers show up¶

• Anomalies page — the largest single driver per anomaly.
• Anomaly detail view — the top five drivers, each with its contribution and share.
• AI summary — the top drivers feed the "why did it happen" explanation.
• Driver tree page — a map of which metrics feed which, built from all the formulas. This is the structure the driver calculation walks.