The open challenge
- Evaluate future possibilities for data analytics at DLA Piper by assessing the content and quality of timecard-transaction data
- Analyse the composition and structure of legal work in Merger and Acquisition deals
How did we help?
The team received an anonymised database extract containing transaction data from the corporate practice group spanning the last 3 years. First, the team mapped the information flow within the database and assessed the quality of timecard recording.
After developing a deep understanding of the database, the team queried 28 tables using SQL to construct aggregated deal variables and key business metrics. The team then visualised the data in R, and engineered new variables, which showed interesting effects on the volume of legal work required for an M&A transaction. These variables included the distribution of legal work activities, and the lawyer-team structure. Finally, with these variables, the team employed k-means cluster analysis (an unsupervised machine learning technique) to detect 8 ‘naturally’ occurring deal types. After further analysis, some of these deal types were discovered to have statistically different margins.
Key facts & findings
– Analysed billing data on 500 M&A transactions over the last 3 years
– Engineered over 100,000 timecards into key deal variables
– Performed k-means cluster analysis to identify 8 natural deal ‘types’
– Discovered certain deal types had statistically different margins
Otto Godwin, Benjamin Gutierrez, Amardeep Jass, Nirbhay Sharma, Shuyu Zhou
Author: Otto Godwin