Background and objectives

An energy network operator has recognised the potential of applying machine learning techniques to predict which of its activities in the field impact customer satisfaction most. The benefit of this prediction helps our client to focus their continuous improvement program. Not only would this focus drive higher CSAT, the regulatory allowance is partly driven by the CSAT score of the network operators, a benefit of 1% of revenue which goes straight to the bottom line.

 

Our approach

The CSAT score is a combination of satisfaction scores provided by customers who have been exposed to repair jobs, replacement jobs and new connection jobs. The satisfaction scores are derived from surveys conducted by a nationwide independent party. Although the CSAT scores are directly related to the overall score the customers give in the surveys, this overall score is not directly derived from the answers the customers give to the multiple questions in the survey. Therefore, to derive decisions to improve CSAT from these surveys, our client recognised the potential of machine learning / predictive analytics techniques to predict which part of for example the end-to-end repair job needs to be improved first as to increase CSAT. We consolidated three types of data to predict which activity has highest impact on CSAT: 1) data from jobs conducted by the field engineers: type of job, time of job, etc.; 2) customer and stakeholder interaction before, during and after the job and; 3) survey data and data on customer complaints and accolades incl. their timing. In close cooperation with the customer services teams and field operations teams, we developed a Random Forest model which predicts which element of the field jobs does affect the CSAT score most.

 

Results

Our predictions showed customers were most unhappy about particular elements of the customer communication plan (in contrast to what our client expected, namely actual activities of field engineers such as street works, reinstatement, etc.). By embedding this predictive model in the stakeholder & communications planning, our client has improved its customer engagement which, among other improvements, has led to winning several customer service awards in their industry and cross-industry. In addition to the broad recognition these awards provide, the financial benefits in terms of additional allowances amount to the full 1% of revenue.