The following example shows how you can optimize the result of a campaign by incorporating RFM analysis with the creation of target groups.

A mail-order company sends its customers its main catalog every year. In addition, the company also sends out different catalogs for specific product groups, but these catalogs are only sent to a select group of the customer base. Other customers have the option of requesting these additional catalogs if interested. Alongside the catalog business, the company also promotes individual products using targeted campaigns directed at selected customers and using different channels: by mail or telephone.

This example involves a campaign to sell a highly profitable knife set. Experiences from similar campaigns performed previously have been gathered and can now be applied as representative campaigns to create target groups and optimize the current campaign. One of these representative campaigns was taken as the basis in SAP BW for creating response rate models for the mailing channel as well as for the telesales channel. These models determine the typical buying probabilities with such campaigns for each RFM segment and channel. Two attributes have been set up in the Segment Builder: one for Buying Probability When Telephoned and one for Buying Probability When Mailed.

You are responsible for defining and optimizing the target group for the campaign.

Proceed as follows:

1. Using the different attributes, first specify a suitable target group from the existing customer master, applying your experience with this type of campaign.
2. You would then like to divide this group into subgroups: one subgroup is to be addressed by telephone, another by mail. You first examine the target group using the attribute Buying Probability When Telephoned. In the Profit Simulation screen, enter the contribution margin for each knife set sold and the cost unit rate for contacting a customer by telephone.

The system uses this information to determine the anticipated profit or loss that would be made if all the customers of the current target group were contacted by telephone. At the same time, the system proposes a filter value that can be used to restrict the target group to those customers that maximize the expected profit. In this example, the proposed filter value is 2.4%.

3. Now create a filter "larger than or equal to 2.5%" for the attribute Buying Probability When Telephoned and use this filter to divide your target group into two target groups.

The first group only contains customers with a buying probability exceeding 2.4% when they are contacted by telephone. This group is then contacted by telephone during the campaign.

4. You now examine the second group using the attribute Buying Probability When Mailed. Enter the cost unit rate for sending an offer by mail.

The system also proposes a filter value for maximizing profit with this target group. In this example, the filter value is 1.3%.

5. Now create the filter "greater than or equal to 1.3%" for the attribute Buying Probability When Mailed. Restrict the second target group to this filter value.

The target group that you started with has now been divided up into a potentially most profitable target group for a telephone campaign and a potentially most profitable target group for a mailing campaign. All customers with a buying probability that is so minimal that profit is unlikely to be made by contacting them by either channel have been removed from the target group. Besides this division into more specific target groups, you receive predictions for the required budget, the number of expected purchases, and the expected profit.

In addition, you can also use the ROI Simulation pushbutton to maximize the Return on Investment (ROI), that is to say, the profit made in relation to the means used.

Since the system calculates in parallel for all possible filter values the effects that each filter value has on profit, ROI, costs, and the expected number of purchases, restrictions relating to budget, channel capacity, or offer availability can also be considered. To incorporate such information, simply select the most profitable value containing the existing secondary conditions instead of the filter value proposed by the system.