Data Screening Tools for Competition Investigations

Data screening tools are empirical methods that competition authorities can use to detect illegal activity. Such methods are often used before opening an investigation, to check for suspicious behaviour that is ‘flagged’ if certain criteria are met, or following a complaint or whistle-blower report, to confirm or validate allegations.

The most common use of screening is aimed at uncovering anticompetitive horizontal agreements, particularly in procurement markets. These are often empirical screens that can be enriched with machine learning techniques. 

In November 2022, the OECD held a roundtable on "Data Screening Tools for Competition Investigations" to explore developments in digital screening tools in academic literature and competition authority practice. The discussion covered cartel screening techniques, the requisite data quality, the value and risks of publicising cartel screens, and investigative successes based on screening.

This page contains all related documentation. 

• Cartel screening is difficult to assess because authorities do not always fully disclose its use. However, interest in cartel screening has increased among competition authorities and some private entities.

• Academic research on cartel screening has expanded significantly, particularly through combining supervised machine learning with traditional behavioural screens. However, supervised methods require training data, so authorities more commonly use simpler screens or unsupervised machine learning techniques.

• Effective cartel screening requires high-quality data and skilled staff. International co-operation can support knowledge sharing, data exchange, and even trained machine-learning cartel screens or datasets.

• Competition authorities primarily use cartel screens to initiate investigations, conduct dawn raids or prioritise cases. Screening results are typically indirect evidence and are usually supplemented by direct evidence gathered during investigations. Rarely, screens have also been used to close cases or support infringement decisions.

OECD note | Note de référence de l'OCDE

Detailed Summary of the discussion EN | FR

Executive Summary with key findings EN | FR