The Competition and Markets Authority (CMA), together with Spend Network, has developed a "Screening for Cartels" tool to help procurers/commissioners screen their tender data for signs of cartel behaviour. The tool was initially tested on data provided from local authorities and recent reports make interesting reading.

Cartels are formed by suppliers working together suppliers with the purpose of maintaining prices at a high level and restricting competition. The forming of cartels to either directly or indirectly fix bidding prices or trade conditions contravenes article 101 of the Treaty on the Functioning of the European Union as well as other national and EU legislation because they distort or prevent competition within the internal market. The CMA estimates that having a cartel in a supply chain can raise prices by 30% or more. This is particularly significant in public procurement procedures, where certain contracts can be issued based on price alone.

To use the tool, procurers/commissioners will need to upload the invitation to tender document, tender responses from all bidders, the identity of the winning bidder and the pricing information contained in each bid. This data is then analysed using a number of specially developed algorithms that identify suspicious signs in three key areas:

  • the number and pattern of bidders;
  • pricing patterns; document origin; and
  • "low endeavour" submissions.

According to the tool's programmers a "low endeavour" submission is one which shows a below average number of edits or that a below average amount of time was spent on its production. The term is used to reflect the lack of effort put in to bids that are designed to be purposefully rejected to create the illusion of competition

The end result is a "suspicion score" for the particular tender exercise. These scores reveal which tender responses are potentially suspicious of cartel activity. Tender responses that produce a high suspicion score do not necessarily prove the existence of a cartel, but it could be used by procurers/commissioner to raise questions about the submitted bid documents.

In a project update report in February 2016, Spend Network claimed that it was able to successfully analyse data from government and local authority procurement information it was given. However the update included the caveat that the tool is still in its infancy and would require a larger dataset before its effectiveness could be fully confirmed. As of July 2017, the dataset had expanded to 100 tenders with nearly 500 bids, thus refining the tool's algorithms.

The report by Spend Network also highlighted the limitations of the tool. Firstly, because of the relatively small dataset that was used for testing, along with the nature of the algorithms used, the tool offers little protection from false positives. For example, a low number of bidders could be a sign of an uncompetitive cartelised market. It could also be the result of a restrictive specification making bidding unappealing for all but a handful of specialised suppliers. Despite Spend Network almost doubling the size of the test dataset from 2016 to help remedy this, it may still be considered as being "small" for the purpose of ruling out false positives. Furthermore, some cartels may operate on an opportunistic basis where the ultimately "losing" suppliers agree to submit inflated prices on a single tender allowing the lowest priced tender to win the competition, and the winning bidder will make payments to the losing bidders. These cartels are likely to be very difficult to identify using algorithms.

One way in which the tool attempts to reduce the instances of false positives and difficult-to-identify cartels is to provide for variable thresholds on long lists of algorithms to help distinguish between cartels and more casual collusion. The weighting of each test can be adjusted to reflect the user's knowledge of the market. This increased level of customisation is designed to help improve the accuracy of the tool.

Despite its relative infancy, the cartel screening tool has successfully proved that algorithmic tests on tender data can be applied to identify suspicious tenders and provides a proactive method for procurers to take to avoid them.

Further information on the tool, including a background report on its development can be found at: https://www.gov.uk/government/publications/screening-for-cartels-tool-for-procurers

CMA website: https://www.gov.uk/government/organisations/competition-and-markets-authority