Search Network

Predicting Demand and Upselling to Clients

 Executive summary

This project applied analytics to identify high-value car rental customers and predict future demand using the company's historic vehicle reservations database. For instance, this allows for the number of vehicles required at a particular location for a future date to be determined.

 

Challenge overview

CarTrawler provides an online car rental solution to over 550 car rental suppliers in 175 countries. The first need of the company was to identify customers who might buy additional items when renting vehicles. The second objective was to improve their existing revenue management strategies by accurately modelling the future demand for vehicles.

 

Implementation of the initiative

Several analytic experts from CarTrawler worked with ICHEC and supplied the necessary information. Interim meetings presented results allowing ensuring the study matched their business needs.

The problem of clean data arose in this project, a common issue in analytics. The cleaned data was used to identify the major customers, the types of vehicles being booked and the major destinations by client segment. This exploration allowed for a number of concrete proposals to be made to CarTrawler, who gave feedback and ranked what it felt were the most important areas to investigate. These were the upselling of additional items within a booking to customers and the ability to model future demand for vehicles.

We explored a number of approaches including k-Nearest Neighbours (kNN), Bayesian Analysis and various Time Series Analysis techniques. These provided the means to tackle the needs raised by the company and to help improve its existing abilities.

"CarTrawler staff have worked with ICHEC to improve the company's consumer behaviour insights and to enhance our forecasting ability by using ICHEC's expertise in data mining and statistical analysis ... we are excited that some of the suggestion made could considerably enhance our revenue stream," say Jason Lambourne, Head of Revenue Management and Analytics at CarTrawler.

Fig.1. An example of the kNN for upselling, if k = 3 the item to be classified, represented by the green circle, will be classified as being part of class A (the brown triangles) whilst if k = 5 it would be classified within class B (the blue squares). (See attachment)

Figure 1 shows how the nearest neighbours classifies customers according to classes which represent purchasing behaviour. Mathematical techniques were used not only for modelling purposes but also for predicting whether customers might be interested in related goods or services. (See attachment)

 

Results and achievements

By applying complex mathematical techniques we have identified improvements to the company's analytical and revenue management processes. What this project has demonstrated is that analytics is a powerful tool for improving the delivery of a company's products or services and it is rapidly becoming an active area of interest in particular for SMEs. The results of this work have encouraged other companies to investigate mathematics and computational science to solve similar business problems.

 

Lessons learned and replicability

The project can be replicated for other businesses such as supply chain management, weather, transport planning, and sports where forecasting or similar analytics techniques are needed to support complex business decisions. From the project's inception it was aimed at solving real business problems and providing definite solutions.

  Technology Transfer
Irish Centre for High End Computing (ICHEC)
http://www.ichec.ie/consultancy
consultancy@ichec.ie

  Jason Lambourne, Car Trawler
http://www.cartrawler.com
jlambourne@cartrawler.com

30 November 2011
print version share on facebook share on linkedin

Executive summary

This project applied analytics to identify high-value car rental customers and predict future demand using the company's historic vehicle reservations database. For instance, this allows for the number of vehicles required at a particular location for a future date to be determined.

 

Challenge overview

CarTrawler provides an online car rental solution to over 550 car rental suppliers in 175 countries. The first need of the company was to identify customers who might buy additional items when renting vehicles. The second objective was to improve their existing revenue management strategies by accurately modelling the future demand for vehicles.

 

Implementation of the initiative

Several analytic experts from CarTrawler worked with ICHEC and supplied the necessary information. Interim meetings presented results allowing ensuring the study matched their business needs.

The problem of clean data arose in this project, a common issue in analytics. The cleaned data was used to identify the major customers, the types of vehicles being booked and the major destinations by client segment. This exploration allowed for a number of concrete proposals to be made to CarTrawler, who gave feedback and ranked what it felt were the most important areas to investigate. These were the upselling of additional items within a booking to customers and the ability to model future demand for vehicles.

We explored a number of approaches including k-Nearest Neighbours (kNN), Bayesian Analysis and various Time Series Analysis techniques. These provided the means to tackle the needs raised by the company and to help improve its existing abilities.

"CarTrawler staff have worked with ICHEC to improve the company's consumer behaviour insights and to enhance our forecasting ability by using ICHEC's expertise in data mining and statistical analysis ... we are excited that some of the suggestion made could considerably enhance our revenue stream," say Jason Lambourne, Head of Revenue Management and Analytics at CarTrawler.

Fig.1. An example of the kNN for upselling, if k = 3 the item to be classified, represented by the green circle, will be classified as being part of class A (the brown triangles) whilst if k = 5 it would be classified within class B (the blue squares). (See attachment)

Figure 1 shows how the nearest neighbours classifies customers according to classes which represent purchasing behaviour. Mathematical techniques were used not only for modelling purposes but also for predicting whether customers might be interested in related goods or services. (See attachment)

 

Results and achievements

By applying complex mathematical techniques we have identified improvements to the company's analytical and revenue management processes. What this project has demonstrated is that analytics is a powerful tool for improving the delivery of a company's products or services and it is rapidly becoming an active area of interest in particular for SMEs. The results of this work have encouraged other companies to investigate mathematics and computational science to solve similar business problems.

 

Lessons learned and replicability

The project can be replicated for other businesses such as supply chain management, weather, transport planning, and sports where forecasting or similar analytics techniques are needed to support complex business decisions. From the project's inception it was aimed at solving real business problems and providing definite solutions.

 Technology Transfer
Irish Centre for High End Computing (ICHEC)
http://www.ichec.ie/consultancy
consultancy@ichec.ie

 Jason Lambourne, Car Trawler
http://www.cartrawler.com
jlambourne@cartrawler.com
 

Attached Documents

PDF icon Irish Mathematics and Industry (Irish_Mathematics_and_Industry.pdf | 1.42 MB)