Uplift modelling
Uplift model, also known as incremental model, true lift model or net model (English), is a method for modeling behaviors (English predictive modeling), the additional effects of a treatment (such as a direct marketing action ) on a individual behavior predicts.
Application
The uplift model is in customer relationship management ( Customer Relationship Management) for up-selling, cross- selling (cross-selling ) and used as a model for customer retention. It was also used in personalized medicine. The Uplift model uses a random control group not only to measure the effectiveness of a marketing campaign, but can predict what change in the behavior of a target person created. It is thus a new data mining technique that is used mainly in the areas of financial services, telecommunications and trade for direct marketing for up-selling, cross- selling, churn and retention.
Measurement of Uplift
The Uplift of a marketing campaign is usually defined as the difference in reactions between a selected group and a random control group. This allows the team Markteting the impact of a marketing campaign to be considered in isolation, and measures the effectiveness of the same. Marketing teams can better budget for a performance- enhancing effect of their marketing campaigns, ie, a result which is higher than that in the control group, can use. The presented below is table contains the details of a hypothetical marketing campaign. There are shown the number of feedbacks and the calculated response rate. For this campaign Uplift of 5 % is achieved in the response rate, that is, there were 50,000 more feedback achieved by the marketing campaign.
Traditional model for feedback
In the traditional model for feedback selected customers are usually selected and tried a predictive model to create that the best position indicator of non- feedback sensors via a number of techniques for prediction model ( predictive modeling) separated. Typically, decision trees or regression analyzes are used. Only the selected customer are used to form the model.
Uplift model
In contrast, the uplift model applies a selected customers and control customers, to create a model to forecast, which focuses on the additional feedback. In order to understand this type of model proposed a basic segmentation, which divide customers into the following groups:
- The normally accessible: Customers only report back if they are approached by a marketing campaign.
- The Safe: Customers who re-register, regardless of whether they are working or not.
- The resistance pattern: Customers never re-register even if they are addressed.
- The Breastfeeding: Customers probably will not re-register if they are approached or even a negative feedback type (eg, a proposed purchase may be)
The only segment that provides the additional feedback is normally accessible. The Uplift model supports a valuation technique which customers can be divided into the groups described above. Traditional models for feedback generally relate to the safe, as these models the normally accessible can not be distinguished from the safe.
Return on Investment
The Uplift model allows to focus exclusively on additional feedback and so can very good returns (English Return on Investment) for conventional promotional activities and customer loyalty activities produce. For example, if only the normally accessible to be included in an outbound marketing campaign, encompassing contact costs and the return per customer can be improved dramatically.
Reduction of negative effects
One of the most effective uses of the uplift model is the avoidance of impacts on customer loyalty campaigns. In the telecommunications industry and financial services customers are often animated in customer loyalty campaigns to extend their contract or insurance contract. The Uplift model allows those customers who would likely not terminated, the silent, remove from the campaigns. Thus, this will not be contacted and therefore not animated to an action (such as a change of provider).
Application for A / B and multivariate testing
It is rarely the case that there is a simple selected group and a control group. Often the selection is a variety of variations of messages or multidimensional contact strategies, which are classified as simple treatment. About B or multivariate tests, the uplift model may help to understand whether the variations in testing obtain a significant uplift compared to other objective criteria, such as behavioral indicators or demographic indicators.