Size of wallet estimation: Application of K-nearest neighbour and quantile regression

Abstract

Size of wallet (SOW) estimation is an important problem to solve from a company's perspective. The total business volume conducted by the customer for a product category across firms is generally unobservable, while the volume of transactions conducted by the customer with the company is mostly accessible. This paper focuses on the estimation of SOW and the estimation of opportunity, which is the difference between the SOW and the actual transactional value of the business that a customer does with a company. K-nearest neighbour (KNN) and quantile regression (QR) are applied here to arrive at the estimations, and their performance is compared. Based on the SOW and opportunity estimates, a company can decide its target segment and design specific marketing strategies accordingly, thereby improving its profitability.

Keywords

• Size of wallet
• Share of wallet
• K-nearest neighbour
• Quantile regression