About This Project

The mechanism is a validated, open-source scoring overlay that sits alongside existing sanctions-screening tools rather than replacing them, and rebalances how much weight jurisdiction-based flags get relative to behavioral-outlier indicators specifically for procurement-corruption triage. Concretely: (1) an ingestion module that pulls beneficial-ownership registry data and matches it against procurement contract award data using the same entity-matching logic as the Tello Arista/Fazekas/Volkotrub 2026 study; (2) a scoring engine that computes standardized behavioral-anomaly scores -- ownership-change frequency z-scores relative to sector peers, entity-age outlier scores, missing-BO-data flags, and political-connection network proximity -- using regression weights replicated from that study's validated fixed-effects coefficients; (3) a blended risk output that keeps sanctions-jurisdiction flags visible (they remain legally required for other compliance purposes) but downgrades their weight in the procurement-corruption-specific score and instead surfaces behavioral-outlier flags as the primary trigger for human investigative review. This is a recalibration tool, not a replacement system -- it does not remove sanctions screening, it stops sanctions screening from crowding out the indicators that actually predict procurement corruption in the validated data. The key design constraint is generalizability testing before deployment: the underlying weights are validated on six European, registry-rich countries, so the tool ships with an explicit confidence interval per country based on how similar that country's registry infrastructure and procurement transparency regime is to the validation set, and it flags itself as experimental/advisory-only in any jurisdiction without a matched local validation dataset. No agency should treat this as a drop-in replacement for judgment; it is a way to stop misallocating investigative attention to a signal that a large-sample study says does not work as advertised, while flagging humility about where it has and has not been tested.

Timeline