Learn how the Centric Data And Analytics team enabled a Midwestern home service company to make more effective, data-based staffing decisions based on machine learning using R for Statistical Computing.
The Business Need
The client (a Midwestern home service company) did not have the capability to approach seasonal staffing decisions using data. Staffing decisions were being made based on intuition. The client was also interested in learning how technician tenure was correlated with technician performance. If the relationship between tenure and performance was strong, it would be profitable to keep more seasonal technicians on staff year round.
Two significant barriers existed:
- The data was not available in a format that could be analyzed.
- The client did not have the tools nor the expertise to perform the data analysis.
Centric’s approach to the work was to use a two-phased machine learning process. The first phase of machine learning focused on the data gathering process. This required organizing and profiling a dataset with the requirements necessary to accurately test a hypothesis. The team applied a flexible data mining approach to overcome uncertainty about which data should be analyzed. This flexible approach enabled the team to maintain a solid dataset that had the simplicity of using filters without a loss of data integrity. After solidifying the dataset and properly testing it for accuracy, the team developed a hypothesis that allowed them to determine if technician tenure was correlated with technician performance.
The second phase of the work was to analyze and build models to test the hypothesis and draw conclusions. Centric’s team of data scientists was able to produce descriptive statistics by plugging the cleansed dataset from phase one into “R” for Statistical Computing. Using this tool, the team built regression models, sensitivity analysis, and decision trees.
A number of different variables were tested to determine impacts on technician performance including location, price, tenure, age, gender, and job title. Using hypothesis testing, the team was able to quickly prove or disprove assumptions. Quick iterations not only allowed the team to learn more about the business but led them to ask new questions and gain new insights.
The Centric team was not only able to get the data in a format which made it possible to analyze, but also provided the expertise to perform this analysis. The quantitative analysis done by Centric allowed them to make informed recommendations to the client. Centric’s data scientists were confident in sharing the insights gained from the analysis. The team also displayed the results with powerful interactive visualizations built in Tableau that made it simple for the client to consume the data.
Centric was able to show the client that the relationship between technician tenure and technician performance was not statistically strong enough to suggest keeping more seasonal technicians year round. This finding held true disregarding various qualities of the technician, location, and price. The client is now able to make informed staffing decisions based on statistical analysis of the data instead of relying on their intuition.
Interested in learning more about this project or Centric’s data and analytics capabilities?
Contact Centric National Data and Analytics Practice Lead, Jeff Kanel.