Predictive Analytics

Predictive analytics covers analysis methods that calculate probabilities for future events from historical data. They use statistical models, classic regression or machine learning algorithms such as random forest, gradient boosting or neural networks. The aim is to take decisions early on a data basis instead of just describing the past.

Typical use cases

In marketing, predictive analytics is central for customer lifetime value forecasts, churn models, next best offer recommendations and lead scoring. In e commerce it is used for stock planning, price optimisation and personalisation of product recommendations. In customer service, models identify unhappy customers early so the company can intervene proactively.

Data quality as a prerequisite

Without a clean and complete data base, predictive analytics delivers no reliable results. The maintenance state of CRM data, consistent tracking implementation and harmonised master data are mandatory. The choice of training and validation data also has a strong influence on model quality, since a model must not be trained on a time period and tested on the same period, otherwise overfitting occurs.

Practical use

In day to day marketing, predictive analytics is useful for directing ad budgets at high value audiences and addressing customers at risk of churn with win back campaigns. In email marketing, send times, frequencies and content topics can be optimised per recipient, so that relevant content reaches the right contact at the right time.