Decionis
Execution Authority · Prevented-Loss Methodology

How we account for the margin loss Decionis prevents.

A CFO should never have to take a vendor's prevented-loss number on faith. So ours is not a marketing figure — it is Governed Execution Value (GEV), computed from live governed decisions and gated so it can never be inflated past what the data supports.

Live snapshot
Early — seed cohort

The public trailing-window sample is currently below our render threshold, so we show an early-seed-cohort status instead of a dollar figure. This is the honesty gate working as designed: rather than publish a number a single large decision could distort, we wait until the sample supports it. Raw counts are available on request.

How the number is computed

1. Price each evaluated action

For every decision that carries a monetary amount — an order's net-margin exposure, a payment, a purchase order, a refund — Decionis records the amount at the moment it is evaluated, before execution.

2. Attribute by verdict

Each priced evaluation is bucketed by its deterministic verdict: authorized (cleared through), escalated (routed to higher authority), blocked (stopped before execution), or delayed (held for human review).

3. Sum the governed value

Governed Execution Value is the sum of those amounts — the economic authority that flowed through the control plane. Prevented loss is the blocked-and-restrained share: value that would have executed without a gate in front of it.

4. Gate the public figure

When the trailing-window sample is below a public-render threshold, the site shows an early-seed-cohort status instead of a number. A dollar figure is only ever shown when the underlying sample supports it.

What we refuse to claim

The credibility of a prevented-loss figure is defined as much by what a vendor won't say as by what it will.

  • We do not attribute a margin percentage or dollar recovery to a named customer unless that customer has verified the number.
  • We do not extrapolate a headline figure from a single decision or a below-threshold sample.
  • We do not count value we did not actually evaluate before execution.