GateHouse
All posts
Case notesApr 23, 2026 · 9 min read

Composing SME credit risk across four sellers.

A walkthrough of one real composition: from a plain-language outcome to 92.4% coverage, 11% overlap removed, and a single automated settlement.

KF
Kevin Flood
Co-founder

Here is one real composition, start to finish. A buyer wants to model SME credit risk in the DACH region. Traditionally that means finding, vetting and integrating several data sources over months. We watched it happen in an afternoon.

The outcome, in plain language

The buyer did not browse a catalog. They stated a job: “model SME credit risk in DACH.” The composition engine read two-sided profiles across the network and proposed a dataset assembled from the sellers who actually move that metric.

composition engine — coverage + overlap readout
Coverage of the requested DACH SME segment, assembled across four sellers, with redundant overlap flagged for de-duplication.

Coverage before commitment

Four sellers surfaced: a telco with the widest reach, a second MNO, a retail-loyalty operator and a POS network. The proposal carried an estimated coverage of 92.4% of the requested segment and a per-seller contribution. Crucially, 11% of the composition was redundant overlap — the same SMEs covered by more than one seller — and it was flagged for removal before a euro was committed.

4
sellers composed
92.4%
segment coverage
11%
overlap removed

Run in the cleanroom

The buyer’s risk model shipped as a container into each seller’s environment. It trained where the data lived. No raw records moved; the boundary counter held at zero rows exported. What returned was the fitted model’s outputs — the computed insight — after passing each seller’s disclosure gate.

One settlement

On completion, GateHouse attributed contribution across the four sellers by the same coverage-and-overlap arithmetic the buyer had seen, and settled automatically — against the buyer’s cloud commit. The buyer paid once. Each seller was paid for unique contribution. Overlap cost no one anything.

From outcome to settlement, the platform did the lifting. The buyer described a job; the network composed, ran and settled it.

KF
Kevin Flood
Co-founder, GateHouse
Talk to the team →

Move permissions, not data.

See how the thesis becomes a platform, on your own estate.