Corporate Business - Data Mining for the Food Industry 

Data mining is the unbiased discovery of non-obvious patterns in a large data set. As applied to the food industry, data mining can be used to both discover and predict customer buying patterns using only existing point-of-sales data.

Knowledge of customer buying patterns can be used to increase profitability through:

• Improved customer satisfaction and market share
• Improved inventory control and category management
• More efficient advertising and sale pricing

  • optimizing loss leader costs vs. profit in increased sales.
  • optimizing advertisement of joint products and product sets

• More efficient product presentation

  • location in store
  • relative location (grouping of items)

• Predictions of which customers will churn to the competition

The power of data mining when applied to an analysis of customer buying patterns is shown in the figure below which displays surfaces of computed conditional probabilities for two distinct customer groups. The surfaces are obtained by computing the probability that the purchase of products in two (different) store departments (the horizontal axes) will lead to the purchase in a third department (the vertical axis). The left-hand panel of the figure shows the results for the store's best group of customers and the right-hand panel of the figure shows similar computations for a group of customers who account for significantly less sales. The differences in buying patterns between the two groups are clear.

Data mining provides the ability to quantitatively distinguish between the buying patterns of the two indicated customer groups and, therefore, to guide a merchant to take focused actions which can transition customers who account for a small amount of sales into the group accounting for the largest volume of sales. CSS Solutions has developed proprietary software to compute such probabilities for an arbitrary number of conditional purchase decisions.