Network Analysis of Communications Data to Support Operational Teams

‍Challenge

‍In many cases, staff do not have dedicated analytical support and have to manually interpret communications data, which is time-consuming and can limit their ability to identify new lines of inquiry. While dashboards can improve accessibility, careful consideration is required when extracting meaningful insights from complex data and making them accessible to users without analytical training.

The objective was to explore how network analysis could be applied to communications data to systematically uncover behavioural patterns and relationships, and to determine how these insights could meaningfully support operational staff.

Solution‍

Butterfly Data collaborated with experienced former practitioners to develop several operational scenarios that showed potential to support casework with network analysis metrics. These scenarios focused on shortlisting important connections within the communications data and aimed to help staff prioritise their caseload or identify new lines of inquiry.

Key features of the solution included:

• Modelling communications data as a graph of phone numbers and their interactions
• A set of similarity metrics designed to quantify behavioural similarity between phone numbers, enabling the identification of those likely to be relevant to a given case
• Node-based metrics to provide structural context, helping staff understand the role a phone number plays within the wider network
• Additional metadata and shortest-path analysis to help staff narrow down their search
• A Python-built network analysis pipeline that automates the generation of insights, presented through a simple Kibana dashboard

The approach was validated through a simulation study using realistic synthetic data from an independent source, and was also applied to a detailed operational scenario developed by experienced former practitioners.

Impact‍

The solution demonstrated that network analysis can effectively uncover meaningful relationships within communications data to support operational workflows.

If operationalised, it could free up staff time by automating the analysis of notoriously complex communications data and give them more detailed insights. While further testing is required, the solution makes a strong case for integrating network analysis into operational workflows, enhancing the value that can be derived from communications data.