Background

  • A government client procuring over a hundred thousand different equipment and materials worth over £800m lacked assurance over its supply chain partners.
  • The client had a trusted group of national and multi-national partners that it relied on to supply goods. 
  • It was impossible to know whether and how contractors operated. However, there was anecdotal evidence that some had merged through M&A activities, creating conflicts of interest and opportunities for fraud.
  • DMRC was approached to develop a data-driven approach to assess and detect any occurrence of procurement fraud.

Approach

  • After an extensive period of interviews and analysis of the inner workings of the clients’ procurement processes, DMRC identified several methods to test for fraud rapidly.
  • Using the clients internal and 3rd party data, we implemented a rapid assessment model, modelling 2-3 use cases per week and analysing the results.
  •  We used several techniques such as network analysis, clustering and outlier detection and apriori algorithms to design and deploy robust detection models.
  • In addition, we also devised probability-based methods based on guidance documents to assess the typical utilisation, lifetime cost and other patterns.

Benefits

  • DMRC identified £80m annually in fraudulent overpricing.

  • Each case of fraud directly linked to a supplier, product and transaction enabling fraud teams to investigate.

  • We produced a comprehensive dataset outlining the linkage of companies as evidence for future investigations.

  • We implemented a regular fraud monitoring tool based on the models developed to proactively identify fraud risk before suppliers were paid.