WIXLIB

Using Analytical Marketing Optimizationto Achieve Exceptional ResultsWHITE PAPER

SAS White PaperTable of ContentsOptimization Defined . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1Prioritization, Rules and Optimization – a Method Comparison . . . . 2Prioritization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4Optimization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5The SAS Approach to Marketing Optimization. . . . . . . . . . . . . . . . . . . 6Organizational Considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7Standard Reports and Sensitivity Analysis. . . . . . . . . . . . . . . . . . . . . . 8Extending Optimization with Business Intelligence. . . . . . . . . . . . . . . 8Balancing Suppression Rules and Constraints . . . . . . . . . . . . . . . . . . . 8Success with SAS Marketing Optimization . . . . . . . . . . . . . . . . . . . . 9Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9Learn More. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

Using Analytical Marketing Optimization to Achieve Exceptional ResultsOptimization DefinedThe complexity of direct marketing has expanded rapidly in recent years, particularlywith the growth of digital marketing channels. Companies today have to make difficultdecisions about targeting the right customers with the right offers while staying withinbudget and channel capacities, all without cannibalizing future sales or saturatingcustomers with too many messages. That is a lot to manage, particularly when multiplecampaigns from one company might also be competing for customers’ attention.Optimization resolves these complex issues by looking at problems in a holistic fashionthat balances the constraints of an organization with the need to improve key metrics.Unlike traditional business-rule methods for allocating campaigns to customers,optimization allows marketers to gain critical knowledge about factors that affect thesuccess of marketing campaigns – such as the impact of adding a new channel, theprobable results of reducing a budget or the consequences of instituting a strategiccontact policy.The best way to explain the differences between traditional approaches and optimizationis through example. This paper will use an example in order to: Show the value associated with taking an optimization approach to directmarketing. Discuss organizational challenges common when implementing optimization. Detail the SAS approach to marketing optimization.1

SAS White PaperPrioritization, Rules and Optimization – a Method ComparisonThe following example illustrates problems that can arise when companies executecustomer-based campaigns where there are limits on which customers are eligible toreceive offers. In cases where one customer only qualifies for one offer, the solution issimple – the customer gets that offer. The problem becomes more challenging, however,when there is a group of customers that qualify for more than one offer.Figure 1 shows this situation. Overlapping sections in the diagram represent customerswho qualify for multiple offers. When optimizing across time periods, the overlap canincrease exponentially. What makes offer allocation decisions even more importantis that customers who qualify for more than one offer are often the most valuablecustomers. Poor decisions about campaign allocation could jeopardize that value.Campaign ACampaign BCampaign CFigure 1: Overlapping sections represent customers who qualify for multiple offers.Companies approach this problem in different ways. In the following example, we willcompare three approaches: prioritization, rules-based and optimization. The first thingmany companies do when attempting to make a decision about offer allocation is todevelop model scores that reflect the probability of response for given customers andgiven campaigns. These model scores, in addition to other values such as the expected2

Using Analytical Marketing Optimization to Achieve Exceptional Resultsrevenue from a response, make up an expected value. Table 1 shows the expectedvalues for each customer-campaign combination. Execution of this campaign also hastwo constraints: Each campaign can be sent, at most, to three customers in the list. Each customer can receive only one campaign.PrioritizationPrioritization is the most common approach database marketers take to solve thisproblem. Put simply, prioritization assigns an order of priority for each campaign beingconsidered within the same time period. For example, it may have been determinedthat Campaign A is the best-performing campaign available. Therefore, it will get itsfirst choice of customers and will choose customers 1, 7 and 9 because they offer thehighest expected values available.Campaign B will get the best customers remaining, and Campaign C will get the rest.Table 1 shows the results of the prioritization method. Shaded cells mark the customerschosen to receive each campaign. Using this approach, the company can expect 655in profit for the three campaigns.By looking carefully at the customers chosen for each campaign, you can clearly seethat there is room for improvement. Specifically, it is logical to think that Customer 1should have received Campaign B, which would have resulted in an improvement of 20. Campaign selection based on this type of reasoning is shown below.CustomerCampaign ACampaign BCampaign 0758656060980110?75 655Expected Return:Table 1: Results of the prioritization method.3

SAS White PaperRulesBased on what we learned about prioritization, the logical question becomes: Why notgive each customer the offer that will result in the most revenue? This question describesthe rules-based approach. This approach establishes rules that look at each customer inorder to determine the appropriate campaign for that customer.In our example, Customer 1 will get Campaign B, Customer 2 will get Campaign C andso on. This seems like a major improvement over prioritization, and in some cases itis. However, the drawback of this approach is that if revenue opportunities exist furtherdown the list of customers, the marketer may not be able to target them because ofconstraints. Remember that each campaign can go to a maximum of three customers.Because of this constraint, Customer 9 cannot get Campaign B, even though it wouldbe a better choice. The rules-based approach would result in a 715 profit for thisorganization.CustomerCampaign ACampaign BCampaign 758656060980110?75Table 2: Results of the rules-based approach.4Expected Return:715 60Improvement:

Using Analytical Marketing Optimization to Achieve Exceptional ResultsOptimizationThe use of operations research techniques enables the best allocation of customers tocampaigns. This method takes opportunity cost into account with the knowledge thatextending an offer to any particular customer could prevent a better offer from beingpresented. Evaluating all combinations simultaneously will result in the best possiblesolution. In this case, a profit of 745 was achieved using the same customers and thesame campaigns. This represents an improvement of more than 13 percent over theprioritization method.CustomerCampaign ACampaign BCampaign 75865606098011075Expected Return: 745 30 90Improvement (business rules):Improvement (prioritization):Table 3: Results of mathematical optimization.While a detailed look at the mathematical methods for optimization is not within thescope of this paper, it is important to note two things. First, this simple example doesnot reflect the enormity of typical marketing optimization problems. Many companiesface similar situations with millions of customers, dozens of campaigns, complexconstraints and sophisticated contact policies. When the scale of the problemincreases, so does the opportunity for improvement. Many large organizations haveseen improvements of greater than 25 percent.Second, the computational power necessary to solve such complex problemstraditionally has been a bottleneck. Intensive research by a team of operations researchscientists and domain experts has yielded a breakthrough algorithm that solves largescale problems efficiently. Due to these innovative approaches, SAS allows marketersto solve these problems in a time frame that is reasonable and flexible enough to fit theobjective.5

SAS White PaperThe SAS Approach to Marketing OptimizationAs mentioned above, any optimization exercise will consist of an objective, a set ofconstraints and a contact policy. SAS Marketing Optimization allows marketers whoknow nothing about optimization techniques to construct a scenario with these threecomponents and then optimize campaigns for execution.Objective – The objective for a marketing optimization problem can be defined in manyways, depending on the overall goals of the campaign. If the overall goal is to increaseprofitability, the marketer can choose profit as the metric to be maximized. SAS providesflexibility in the goals of the campaigns so that the optimized value can be the result ofan equation of two or more metrics. In other cases, the marketer might set an objectiveto minimize, such as risk or cost.Constraints – Constraints enable marketers to specify key marketing limits such asminimums or maximums for spending. Constraints can also be set at the customersegment level. Such constraints can involve: Budget. Set the budget constraints for any or all campaigns. In addition, budgetconstraints can be created at the individual communication level. Cell size. Very often, campaigns need to be a certain size to be worth executing.Marketers can create constraints that reflect the real nature of the direct marketingworld through minimum or maximum cell sizes. Channel capacity. Outbound and inbound channels often have limits, whether interms of the total hours a call center can handle or the number of pieces a mailhouse can send out. Custom. Constraints can be constructed such that they enforce a variety ofspecific limitations. For example, geographic constraints may dictate that acertain number of customers are contacted within a certain region. There maybe additional constraints that ensure a proper ratio of high value to low valuecustomers are contacted across campaigns. ROI. All campaigns can have an additional constraint that drives toward athreshold so that a certain ROI is targeted across the campaigns.Contact policy – Contact policies are important for planning the number of allowabletouches that the overall campaigns or brand can have on each individual customer.These can be set in a variety of ways: Maximum contacts. A limit can be placed on the number of touches percustomer for the overall cycle. For example, an organization might say that eachcustomer can be contacted only twice per cycle. This can be maintained at theoverall level or the individual customer level. Group/subgroup. Contact policies can be constructed so that they allow certaintypes of communication more leeway. A credit card company may want to limit theamount of a certain expensive offer, for example.6