AutosAndBikes (the name was changed) is a leader in the USA Automobile resale market for cars, bikes, and ATVs. They purchase and resell used automobiles from individuals through an online portal. Experts evaluate the market and quote the buy price (RFQ) based on the expected sell price. That manual evaluation was error-prone, so AutosAndBikes wanted to automate the RFQ pricing using Machine Learning while keeping their business interest safe.
1. The volume of RFQ (1500+) was difficult to handle manually resulting in a costly backlog.
2. The manual inspection and quotation process were non-transparent and error-prone, which further exacerbated costs.
3. Many automobiles remained unsold due to unfavorable pricing and had to be liquidated in auctions.
1. We assembled historical records, transactions, and other unstructured data into one coherent dataset.
2. We cleaned data for completeness, inaccuracies, and outliers.
3. Together with the client we modeled the factors that affect the automobile buy price.
4. We automated the RFQ process with the following features:
a. The machine learning model for buy price prediction.
b. Algorithm to indicate the need for expert advice (5% of cases).
c. Cloud API deployment and integration with the existing system.
After our intervention, the client was now able to:
1. Predict the actual sale price of their cars with an accuracy of 98%.
2. Allocate the expert advice to fewer cases (recommended by the algorithm), freeing up valuable expert resources.
3. Reduce the size of the team handling RFQs by 40%.
