Dissertation Series - Data Mining Tesco Clubcard Case Study

In 1994 Tesco piloted their Clubcard scheme, then went on to launch it countrywide in 1995 (Humby et al, 2004 p.14), although loyalty schemes were also being operated by other retailers Tesco intended to make further use of the data generated as a result of it.

Safeway had tried this with their ABC card which they later abandoned and had suffered with the too much data issue having compared it to “drinking from a fire hose” (Humby et al, 2004 p.99).  This scenario of too much data was a common theme throughout the research carried out in the literature review and often led to the use of data mining.

Dataset content

A huge amount of data was being created as a result of the Clubcard scheme; till transactions broken down to product level and attributing to the Clubcard holder that made the purchase (Humby et al, 2004 p.96).

Dataset analysis

Tesco outsourced the analysis of the data to a company called “Dunnhumby” as they did not possess the IT skills or infrastructure inhouse (Humby et al, 2004 p.96).  In addition technical limitations of the time meant that Dunnhumby weren’t able to process all of the Tesco data (Humby et al, 2004 p.97), in fact in excess of 50 million transactions (shopping trips) were held in the first 3 months of the scheme (Humby et al, 2004 p.96).

Dunnhumby took the approach of analysing 10% of the collected data and then worked to apply what had been learnt to the entire dataset (Humby et al, 2004 p.97).

Dataset quality

Errors contained within data can cause false positives in data mining (Thrasingham, 1999 p.93) and Dunnhumby encountered this with the Clubcard data having multiple users of one card, users holding multiple cards (in the case of loss/theft) or even local issues preventing customers reaching the store for a period of time (Humby et al, 2004 p.98).  However as the data was collected by Tesco hardware automatically, there was little likelihood of missing data or inconsistent data issues to be encountered.

Resistances

Whilst the majority of the Tesco case study talked about positive feedback (Humby et al, 2004 p.116), there were periods during which there were resistances to the processing of personal data (Humby et al, 2004 p.177).  During 1997 there were 20 complaints made to the Data Protection Registrar (which was the precursor to the Information Comissioner), these complaints pertained to the use of the data collected for the clubcard scheme that was subsequently used in a Tesco Personal Finance mailing campaign (Humby et al, 2004 p.177).  As a result of the complaints and subsequent meetings with the Data Protection Registrar Tesco revised their practices to not pass details from clubcard to third parties (Humby et al, 2004 p.179).

Benefits

Tesco realised a large number of benefits from the Clubcard scheme, with customers feedback being that the targeted mailings were viewed separate to other commercial mailings (Humby et al, 2004 p.116). 

Tesco were also able to more accurately target marketing based on customers buying habits and as such get a higher return, this marketing was coupon based mailing which results in customers receiving a bespoke combination of coupons based on their buying patterns (Humby et al, 2004 p.117).

Summary

Tesco encountered technical issues at the beginning of the club card project, mainly caused by the technical limitations of the time.  This resulted in them outsourcing the project to a third party, who then only analysed a small sample.

Tesco encountered resistance to their sharing of personal data with third party marketers during the Tesco Personal Finance mailing, although changed their approach to data sharing and mailing. 

Otherwise the Tesco viewpoint was that the project was hugely successful and well received amongst its customers.