EVIDENCE · PROVENANCE · LIMITS

Not every number has the same evidence status.

SettlePoint separates operational software, development estimates, sealed-confirmed risk results and client-specific executable evidence. This page says which is which.

THE EVIDENCE CLASSES

Four tools. Four precise claim ceilings.

Quote Margin

Operational for supported cells. Development estimate of adverse midpoint movement; not an executable quote or guaranteed client protection level.

Margin Watch

Live measured headroom plus a development estimate of remaining midpoint path risk; not proof that a warning prevented a loss.

Managed Spot

Operational execution workflow with a sealed-confirmed risk result: Risk First dispersion was 20%–26% of unmanaged due-date settlement.

Bank Check

Deterministic comparison. It becomes client-specific hard evidence when like-for-like executable quotes or actual fills are supplied.

DEVELOPMENT ESTIMATES

Useful for control. Labelled before they are used.

Quote Margin: 121,230 causal development predictions produced 20 supported pair, side and horizon curves; 10 cells are withheld. Mean historical midpoint coverage was 80.38% for Lean, 89.57% for Balanced and 95.04% for Strong.

Margin Watch: 134,616 causal development predictions produced 19 supported curves; 11 cells are withheld. Mean historical midpoint path coverage was 86.15% for Lean, 94.44% for Balanced and 97.57% for Strong.

Those means summarise supported cells. They are not confidence statements for one invoice. Every live result must carry its pair, importer/exporter side, horizon, target and supported/withheld status.

SEALED MANAGED SPOT RECORD

The strongest approved statistic is risk, not average return.

Managed Spot mechanics were frozen and graded on 51,517 invoices from periods not used to design them. Across four sealed party-era cells, Risk First compressed outcome dispersion to 20%–26% of unmanaged due-date settlement.

Balanced and Opportunity First were positive in approximately 69%–75% of 16 calibration quarters, but results varied by period, pair and client side. That is historical calibration, not an expected return and not a promise that waiting will beat settling now.

HONEST LIMITS

What these numbers cannot do.

Historical midpoint bars are not executable bid/ask RFQs. They do not contain actual quote expiry, bank markup, slippage, payment fees, value-date adjustments or the client’s realised all-in settlement price. A modelled warning is not a saving by itself. Hard value is counted from comparable executable bank quotes and realised spot results against the inception settle-now rate, net of disclosed costs.

SettlePoint does not combine the headline results of Quote Margin, Margin Watch and Managed Spot into one ROI number. That would double-count risk and value. A client case starts with invoice, RFQ and fill evidence and is then tested prospectively under frozen rules.

Judge us on the record.