Decisions

Every verdict Straja has made from real product usage, backed by an append-only record on disk. Updates automatically as new decisions land. This is the operating log, not the calibration run: the 75-of-75 figure quoted elsewhere is 15 labelled scenarios scored 5 times each, which is a different measurement on a different set.

live

Total decisions

PASS

SOFT

INTERVENE

€0

Amount held under INTERVENE

across 0 checkout holds

—

Override rate

proceeded after the hold · 0 cancelled outright · 0 left unresolved

Modelled projections — not measurements

Every input below is either measured here or a stated assumption, and each figure says which. Nothing here is a measurement of real-world production behaviour.

Measured here — average amount under an INTERVENE hold, €0, across 0 real checkout holds.

Assumed — 10,000 checkouts/month, a 2% INTERVENE rate, 30% of holds being genuine fraud, 5% abandonment among people held, and €0.125/transaction. All five from business-case §4.

For reference, this suite currently produces 0% INTERVENE across 0 checkout decisions. That is calibration data, not traffic, and it does not drive anything above — the fixture mix decides it, not real behaviour. Still under 20 decisions, so it moves with every new run.

€0

Annualised protected value

200 holds/month (10,000 × 2%), of which 30% are genuine fraud, at the measured €0 average hold. That is NaN% of the partner's annual volume — on-ramp fraud is cited at roughly 0.5-2%.

60 vs 10

Frauds stopped vs checkouts lost, per month

Of the same 200 holds, 30% are genuine fraud stopped and 5% of the held abandon the purchase. A SOFT check is one acknowledged line and isn't counted — only the 24-hour hold carries real abandonment risk.

€1,250/mo

Projected Straja revenue, this partner

Assumes 10,000 checkouts/month at €0.125/transaction (business-case §4 list price) against a €0.044/transaction weighted API cost — ≈65% gross margin.

All decisions (0)

No decisions recorded yet.