Analytics

Client challenge

A Salvage Car client wanted to use a scientific, data driven methodology to set reserve prices for their salvage cars during auctions. The reserve prices were being set based on the “gut” feel of insurance company executives meaning revenue loss and/or unsold inventories. The data required to create price prediction models was unclean, un-integrated and messy. There was no Machine Learning deployment or Life Cycle Management strategy in place.

How we made value

Sogeti conducted multiple workshops with key experts to determine the existing process of reserve price setting and the probable variables involved in making the decision. We shortlisted these key predictor variables and performed data profiling, and analysis to determine distributions, outliers, missing data etc. Using this info we defined and implemented methods to manage outliers and impute missing values using a variety of machine learning algorithms and deployed them on Azure ML Services.

Sogeti impact

We ensured a reduced inventory up to 10% and maximization of revenue. Our data driven method of determining sale prices helped in setting an optimal base price and we provided a feedback loop for continuous improvement, including a dashboard to help determine when the model results drift from reality.