Campaign Optimization using RFM Analysis 
Campaign Optimization using RFM Analysis
Purpose
RFM analysis is an empirical procedure that has long since been successfully applied in database marketing and that is used to make predictions about response rates for campaigns and to optimize campaigns. In this way, RFM analysis is a tool for improving the profitability of campaigns or for optimizing the ROI. For the practical implementation of this procedure, SAP delivers the analytical application Campaign Optimization using RFM Analysis.RFM analysis is applied during the preparatory phase of a campaign, helping you to identify the optimum target group. Target groups are frequently established by focussing primarily on demographic criteria. RFM analysis adds to these a criterion based on customer behavior - the predicted probability of a response.
The predicted response probability is drawn from the analysis of past data about customer behavior. This typically involves data about sales orders made by customers. In most cases, the response rates for forthcoming campaigns can be predicted with a high level of reliability on the basis of this kind of behavioral data together with experience gained from the responses obtained in comparable campaigns in the past.
Data about customer behavior is used to determine the values Recency (R) (how recent the last purchase is), Frequency (F) (how often purchases are made), and Monetary Value (M) (the amount spent by a customer), and these three values are applied to divide customers into segments. If the data used is taken from past sales orders, for example, then the R value is the date of the most recent purchase prior to the key date of the analysis, the F value is the number of sales orders made, and the M value is the financial amount represented by the sales orders for a specified period up to the key date of the analysis. Along with sales orders, you can also draw on other behavioral criteria, such as calls received or inquiries made by post. To use RFM analysis most optimally, it makes sense to experiment with the different options appropriate to your field of business. The system also allows several approaches to be used in parallel as
RFM segmentation models.In segmentation, the system assigns each customer to a specific
RFM segment. R is considered the factor having most influence on the predicted response probability, whereas F is considered less influential. In segmentation, the M value is assumed to have least influence on response rates. This order of importance is most frequently used, and its appropriateness can be checked using the delivered queries.The effectiveness of the RFM analysis is enhanced by linking it to data relating to experience with similar campaigns carried out in the past. An RFM segmentation of the customer base is made retrospectively for the time when these campaigns were launched and the actual response rates for the customers addressed are measured per RFM segment. If these campaigns are considered representative, then experience shows that these response rates are typical for the corresponding RFM segments and can therefore be applied in the prediction of subsequent campaigns. "Representative" does not merely mean that the type of campaign has to be sufficiently similar, but - more significantly - it also means that the customers addressed in these previous campaigns were sufficiently representative. This is most easily achieved by creating a target group from a random selection of customers in your customer master. You can then perform a campaign using this nonspecific yet representative target group. This usually generates high costs initially. However, applying past campaigns with nonrepresentative target groups can lead to distortions that can severely hamper the RFM analysis or indeed make it totally ineffective. Performing a representative campaign should therefore rather be seen as an investment that you can soon offset with the measurable increase in profitability of future campaigns.
The main application of this procedure resides in predicting response rates for a given target group and simulating the profitability and ROI for campaigns addressing those target groups. You can use the insights gained with RFM analysis by limiting, for example, the target group to all customers with a minimum response rate. This enables you to increase profitability because the costs of addressing customers with whom the probability of a response is particularly low can exceed the anticipated profit from their responses. It is also conceivable to split target groups into more specific ones that you can then address individually using different channels of varying costs.
Even though RFM analysis can be used in many situations, the assumed relationship between the measurements of past behavior and the probability of a response in future campaigns may not be sufficiently strong in your field of business, or it might be debatable whether such a relationship even exists in your field of business. To establish the appropriateness of this procedure, use the options proposed by SAP. For information about these options, see the documentation in the SAP Library for SAP BW under Business Content ® Analytical Applications ® Customer Relationship Management ® Campaign Optimization using RFM Analysis ® Checking the RFM Analysis.
Prerequisites
RFM analysis can only be applied to customers for whom you have behavioral data. For the definition of target groups, however, it is common practice to also use customer addresses taken from external providers, even though there is no behavioral data for these customers. Nevertheless, RFM analysis can still be used even if there is no behavioral data for some of the customers in the target group. Only those customers in the target group with such data are considered during valuation and optimization. In such instances, the potential for optimizing the campaign is reduced accordingly. Nevertheless, the result is still an optimized campaign.
If you would like to use the RFM analysis functions provided by SAP, you need the following software components in mySAP.com:
In SAP CRM online, RFM analysis is integrated in the Segment Builder and allows you to predict the response rate as described above as well as to simulate profitability and ROI with the different target groups you have determined.
In SAP BW, the options for determining the RFM segmentation and the RFM response rates are combined in a single transaction for RFM modeling. You can then apply the results determined during RFM modeling elsewhere outside of the CRM Segment Builder.
Process Flow
The steps listed here provide a general overview of the process flow for an RFM analysis:
For more detailed information about performing a campaign using SAP CRM, see
For more detailed information, see the documentation in the SAP Library for SAP BW under Business Content ® Analytical Applications ® Customer Relationship Management ® Campaign Optimization using RFM Analysis ® Segmentation.
For more detailed information, see the documentation in the SAP Library for SAP BW under Business Content ® Analytical Applications ® Customer Relationship Management ® Campaign Optimization using RFM Analysis ® Response Rate Calculation.
This provides you with the information about which customer currently belongs to which RFM segment and what the response rate is for that RFM segment.
For more detailed information, see the section
Using RFM Analysis in the Segment Builder.