Statistical Consultants Ltd

Data Mining Services

Data mining involves using statistical and computational techniques to reveal useful information (usually) from a large amount of data.

When applied to data from a sales database, data mining techniques can reveal valuable insights into customer behaviour.

Data mining techniques / applications provided by Statistical Consultants Ltd include (but are not limited to) the following:

Predictive Modelling of Customer Value

Some customers/clients are more valuable than others. Time and money is often wasted on advertising and soliciting attempts on people unlikely to spend. It may be possible to estimate a customer’s propensity to spend, using statistical modelling techniques.

Example – Mail Order Catalogue:
A company would like to send a mail order catalogue to customers listed on their database. The following steps could be taken to reduce wasteful spending:

The catalogue would be sent out to a small sample of customers from the database, as part of a pilot study.
If a sampled customer orders something from the catalogue, all information related to that order would be recorded.
A statistical analysis would be performed that estimates the mathematical relationship between various characteristics of a customer and their value to the company. For example, a customer that has bought items A, B and C in the past may be highly likely to order item D from the catalogue.
The statistical model would then be applied to all other customers in the database, allocating an estimate of worth to each customer.
The catalogues would then be sent only to the customers the company is likely to profit from.

Classification of Customers / Clients

Linear Discriminant Analysis (LDA) for classification

Classification techniques are a type of predictive modelling that predicts categories rather than quantities.

Classification models could be used to classify new customers / clients into known groups. For example: Predicting whether or not a loan applicant would default based on their income, their credit ratings, the amount they borrow, and other data about the applicant.

Cluster Analysis / Market Segmentation

Cluster analysis involves forming groups of similar observations. Once the main clusters have been identified, the statistical characteristics of each cluster would be investigated.

In a marketing context, this usually involves grouping customers/clients. This is known as market segmentation. By finding the main types of customer/clients, a business can then gear their marketing, pricing and production strategies towards each group rather than take a ‘one size fits all’ approach.

Customer Retention

Efforts made to retain existing customers tend to be cheaper than advertising to attract new ones.

If a sales database contains the records for individual customers, data mining could be used to investigate which customers appear to have stopped buying and why they have stopped.

This information could then be used to help develop a customer retention plan. Customer retention is also known as customer churn prevention.

Product Linkage

It may be common for two or more products to be sold together. Once the product links are identified, the linked products could then be marketed more effectively by having them marketed together. Marketing strategies that take into account product linkage include:

Specials e.g. loss leaders, bonus gifts, joint discounts etc
Up-selling (especially in an online setting)
Reorganising shelves and catalogues to improve the chances of the linked products being seen together

There are several statistical approaches that can be taken in a product linkage analysis, including:

Probabilities e.g. if someone buys items A, B and C in a single order, the probability that they would also buy item D in the same order is 0.87.
Cluster Analysis e.g. there are many orders that have an item combination similar to the combination of items A, B, C and D.

Other Data Mining Applications

In addition to marketing purposes, data mining techniques can be applied to:

Fraud detection e.g. detecting accounting fraud
Finance
Agriculture
Education
Molecular biology / bioinformatics
and more