Abstract
Data quality is usually scored in isolation. GateHouse argues quality is intrinsic × extrinsic: a dataset’s own properties multiplied by its position in a live network of demand, benchmark and linkability. This paper defines both factors and how they combine.
The single-tenant blind spot
A dataset can be pristine and worthless, or messy and indispensable. Intrinsic quality tools — profilers, catalogs, lineage — cannot tell the difference, because value is relative to a market they cannot see. Network-relative benchmarking is precisely what a single-tenant tool cannot produce.
Intrinsic signals
Auto-profiled on connect, inside the container: fill rate, freshness, width, consistency, coverage and stability, each with drill-downs. No taxonomy authoring; the profiler does the classification.
Extrinsic position
The extrinsic factor stitches demand, pricing and linkability onto the intrinsic profile: how a dataset ranks where it will actually be sold. A dataset in the top decile of demand is worth repricing even if its intrinsic score is unchanged.
Linkability
A dataset’s value multiplies with what it can be joined to. Schema- and taxonomy-based linkability (IAB Taxonomy, SDA, DataLabel heritage) scores how composable a dataset is with the rest of the network — the raw material of coverage and overlap estimation.
Gap analysis
Quality becomes actionable when remediation is priced: “raising fill on this column moves you from the 40th to the 12th percentile and is worth an estimated €X in demand.” Sellers prioritise by value, not by tidiness.
Why it compounds
Every transaction sharpens the linking graph, which sharpens extrinsic scoring, which improves the next composition. Quality-as-position is not a static grade; it is a live read that gets better as the network transacts.