Hedonic pricing models for used items

Abstract

Purchase of used items has grown rapidly in India. According to a recent study done by Olx1 , urban India holds more than 22000 Crore of items with unlocked value lying idle in their homes. Used car market is one such category of used items, which has rapidly grown in India over last 5 years. Third party verification vendors stand to gain from the large size of used car market, high per ticket price of purchase and the need for verification vendors to establish trust and facilitate transaction. One of the key problems is to find a good estimate of price. Pricing methodologies are still in nascent stage in India. In this project, we have demonstrated a methodology to build a pricing engine for used cars based on list prices available in market. The methodology consisted of three parts: 1) Cleaning the data, 2) Exploratory Data analysis 3) Building models. After cleaning and preprocessing the data, we narrowed from 38000 initial data entries to 7150 data entries. Exploratory data analysis helped us get an idea about potential key variables. We then developed models for three cars: Maruti Swift, Honda City, Tata Indica/Indigo. We have also developed models for two other car categories: Maruti and Hyundai, as there were many models by the same manufacturer but most of the individual numbers are very small. From the models, we realized that there is an error of 18% for Honda city, Tata Indica, Hyundai models and Maruti Suzuki models, and 10% for Maruti Swift. This clearly shows that the current pricing engine has significant scope of improvement. By including more parameters such as number of owners and condition of car, the accuracy of model can be further increased. Using the methodology demonstrated above the company can develop pricing models for different cars. Further, they can update them regularly based on a systematic approach and will benefit from increased business opportunities.