Prior Situation / Scenario:

  • Intuition as guide to start debt recovery process
  • Losses due to failed debt recovery processes
  • Regulatory changes made the process more expensive
  • Active Alternative Fake Energy providers on Social Networks

Client Challenges:

  • Create an objective, data-based solution to decide when to start debt recovery process
  • Prioritize customers based on their probability to pay and the expected value to be recovered
  • Reduce energy theft

Strata Solution/ Key Enablers:

Development of an indicator that combines:

  • A machine learning algorithm to predict the likelihood of an account to pay after fraud was detected.
  • Attributes that are related to higher fraud debts.
  • We develop an automated Social Network analytics scrapper to identify suspicious alternative fake energy providers.


  • We enable the ranking of customers that had a higher propensity to fraud.
  • We Improved the fraud management process efficiency
  • Timely debt recovery process improvement
  • Early detection of “Fake Energy providers”


With the application of the machine learning algorithm to the debt recovery process, the company increased the amount of debt repaid by 19% and reduced losses.