Case Studies


Case Study: Predicting Mean Time to Repair - Utilities

This customer needs to estimate repair time more accurately to properly set customer expectations.


Company   Utility Company - Name Withheld
Product   Power
Location   Eastern USA

As soon as the power goes out, people are calling not only to report it, but to find out when it will be restored.  Having a good estimate is an important element of customer service in these situations. Estimates were "about 20 minutes" too often.


Objective   Find key drivers and develop more accurate models that predict Mean Time to Repair (MTR) within a given tolerance.
Method   Approximately 100 variables in a history of more than 5,000 cases were modeled to estimate MTR based on the situation at hand.  This included the time, place, day of week, equipment involved, crews available and many other factors.  This data was modeled using BioComp's Process Modeler software to find the most accurate model..

With a quick application of Process Modeler's modeling technologies we were able to increase the estimate accuracy from only 7% of the cases within 20 minutes of actual repair time to 32% of the cases within 20 minutes of actual, a 466% improvement.  The first chart shows the customer's current estimation accuracy (vertical axis is MTR, blue is actual, red is the estimate):


Estimate Using Customer's Current Method

MTR Original


The predictive model (shown BELOW) shows a CLEAR improvement (red is actual, blue is estimated by the model):

Predictive Model's Estimate



Reference Available?