Purchase Behaviour Users Segmentation

Client: was the major worldwide provider of online auctions service.

Needs: carrying out an e-mailing campaign targeted to groups of homogeneous users in terms of their historical purchase behaviour (users who tended to show a similar behaviour in the past 24 months).

The exhibition organiser needed access data through a Web application to:

• Manage the database;
• Easily extract data from the archive/database;
• Develop/build statistical reports;
• Track activities on contacts.

Project: the project has been carried out by Freedata together with a marketing agency who designed the e-mail layout and content for the campaign. Descriptive statistics as well as multivariate statistical techniques were necessary in order to fulfil the project objectives. The project that was executed in two steps: online auction users were first classified into three groups according to the purchase behaviour showed in the previous 24 months. Purchase behaviour was defined in terms of: Value (average value of a purchase), Frequency (average number of monthly purchases), Class (attitude to change the kind of objects purchased during the considered time interval).The analysis led to the identification of three large groups of users:- Believers: users showing a constant, high value of purchases;

• Persuadable: users whose purchase behavior varied during the considered period. Their purchases value was medium/high. These users were considered to be potentially more receptive to a more refined communication;
• Resistant: users showing a constant, low value purchase behavior.

People identified as Believers or Resistant received a targeted e-mail. A more refined segmentation of the individuals classified as persuadable was performed (persuadable users were considered to be the ones most likely to be seduced by an even more refined e-mailing campaign). A multivariate statistical analysis was carried out, the outcome of which led to the identification of 8 clusters characterized by homogeneous characteristics. The variables used to detect these groups were:

• Age
• Gender
• Average purchase value
• Average monthly frequency of purchases
• Categories of products purchased
• Number of categories of products purchased
• Behavior showed during the auctions (in terms of number of offers made, timing etc.)

Results: the analysis led to the classification of each client according to the distinction made among Persuadable, Believer and Resistant Users./ Each client belonging to the list provided by the customer was identified with a label indicating his/her classification as believer, persuadable or resistant. In addition, each persuadable user was assigned to one of the eight clusters detected.

Timing: 6calendar weeks.