Data analysis and interpretation

Computers are invariably used to carry out the quantitative analysis of questionnaire data. Basic marketing analyses can be carried out using such software analysis packages as SNAP or MARQUIS on a personal computer. More sophisticated analyses can be conducted using packages such as SPSS-PC and NUD.IST.

Basic analysis of questionnaire data may be at the descriptive level (e.g. means, frequency tables and standard deviations) or on a comparative basis (e.g. t-tests and cross-tabulations). More sophisticated analysis may search for relation­ships (e.g. regression analysis), group respondents (e.g. cluster analysis), or establish cause and effect (e.g. analysis of variance techniques used on experimental data).

The increase in the amount of data sourced via direct marketing activity and loyalty cards has led to a need to analyse this data so that meaningful purchasing information and trends may be observed. The analysis of this data is called 'data mining'; e-Marketing 4.2 discusses developments in this field.

When interpreting marketing research results great care must be taken. One common failing is to infer cause and effect when only association has been established. For example, establishing a relationship that sales rise when advertising levels increase does not necessarily mean that raising advertising expenditure will lead to an increase in sales. Other marketing variables (e.g. salesforce effect) may have increased at the same time as the increase in advertising. A second cautionary note concerns the interpretation of means and percentages. Given that a sample has been taken, any mean or percentage is an estimate subject to 'sampling error', that is an error in an estimate due to taking a sample rather than interviewing the entire population. A market research survey which estimates that 50 per cent of males but only 45 per cent of females smoke do not necessarily suggest that smoking is more prevalent among males. Given the sampling error associated with each estimate, the true conclusion might be that there is no difference between males and females.

4.2 e-Marketing

Data Mining

The increasing use of information technology to capture data about customer purchase behaviour has, in recent years, provided companies with mountains of new data. Consider, for example, the data that is held by just one supermarket through the analysis of the loyalty cards it issues to its customers. Clearly, such information has the potential to be a rich source of new knowledge. This only becomes a valid objective, however, if ways can be found to 'mine' these veins of knowledge.

This problem has attracted the attention of a number of software companies. The outcome has been the development of a new generation of data-mining tools from companies such as AbTech (www.abtech.com), Information Discovery Inc. (www.datamining.com), NeoVista (www.neovista.com) and Ultragem (www.ultragem.com). Their tools permit the conversion of data 'warehoused' within the firm into predictive information on:

• classifying data into meaningful entities for determining interactions between market variables;

• modelling explicit relationships between datasets;

• clustering data into understandable market categories or segments;

• determining deviations that may be indicative of new trends in customer behaviour;

• forecasting whether patterns emerge, leading to a better ability to predict the future;

• sequencing-that is, seeing if one event will lead to another (the purchase of a barbeque, say, will lead to the purchase of a picnic table or related products).

The result is that those companies that have adopted data-mining systems have a much more detailed understanding of customer behaviour. This knowledge can be used for cross-product selling, making offers to enhance customer loyalty and identifying new ways of further improving the quality of customer services.