Preparing for future: Forecasting cab booking trends for different booking methods

Project Details

Term: 

2013

Students: 

Nikhil Gupta, Deepti Singh, Shriya Shekhar, Piyush Gupta, Piyush Singh

University: 

ISB

Presentation: 

Report: 

Our client Yourcabs.com operates a platform to efficiently connect consumers in need of transport, with vendors in need of increased occupancy. Customers can make bookings through 3 possible methods- Online, Mobile website and phone. While traditionally most of the bookings have been coming from the phone method, the other 2 methods are fast catching up in terms of no of bookings being made. To maintain infrastructure to cater to a certain demand level in any of the booking methods, certain fixed and variable cost is incurred by our client. As such, matching demand with optimal infrastructure for each booking method is the first step in minimizing these costs. Various costs associated with different booking methods are given below:
Phone: Maintaining operators and phone lines. This is dictated by the client management agreed customer service levels for each demand point.
Online Booking: Dedicated servers needed to cater to the online traffic. A good forecast on an hourly basis can help us to maintain a base level of infrastructure and then rent optimal server capacity to cater to any predicted peaks.
Mobile Website Booking: Dedicated mobile servers to cater to the mobile traffic. As with online bookings, a good forecast can help to maintain a base level of infrastructure and then rent optimal server capacity to cater to any predicted peaks.

There are asymmetric costs associated with demand forecasts. Over forecast will lead to excess spending in capacity infrastructure while an under-forecast will lead to customer dissatisfaction, which might lead to lost sales in this competitive market. This loss is increased manifold if we look at the customer lifetime value (LTV).

This report aims at minimizing the tradeoff by maintaining optimal infrastructure for each method. This is done by forecasting future demand through each method by making use of available 3 years data.

Application Area: