Optimizing Sales Promotions

»Buy two, get one free« or »Valentine’s Day Special Edition« are familiar images at the supermarkets. Taken together, price discounts, special packaging, displays, giveaways, and other actions to promote a product comprise an area called consumer-focused trade promotion.

Optimizing Sales Promotions

Modeling with the Help of Smart Data

»Buy two, get one free«, »30% more content«, »Valentine’s Day Special Edition« are familiar images at the supermarkets. Taken together, price discounts, special packaging, displays, giveaways, and other actions to promote a product comprise an area called consumer-focused trade promotion.

The manufacturer must decide what products to pair with which of these sales driving measures so the use of the budget will result in the biggest profit. Product managers have the job of not merely selecting an appropriate tactic for an individual product; rather, they have to build a portfolio of measures for various products.

Demoansicht eine Promotionskalenders
© ITWM

Demo view of a promotion calender

Cannibalization Effect: »Smart Data« succeeds

The main targets are costs, total sales, and volume. Each measure produces not just an increase in sales, but also can have a negative influence on the sale of other owned brands. This effect is known as cannibalization. For example, an increase in the sales of nut chocolate at a reduced price can mean fewer sales of milk chocolate at the regular price.

The sales forecast for a trade promotion is one of many project examples in which smart data – the linking of framing models with calibration data – has proven itself superior to the pure, data driven, Big Data models.

This is particularly due to the fact that the available sales data provided from various sources does not always have the quality required to be sound for both the base model and the exception (response to the promotion measure). The underlying model resolves this problem by providing control over the forecasts to ensure they are always plausible, or that otherwise data quality problems are exposed.