In a discussion on the value of segmentation, my friend Janet asks, "Where's the lift?" When management asks this question, are you prepared to give an answer?
Lift measures the effectiveness of a marketing campaign sent to targets selected by a model as compared to one sent to a randomly selected list. Questions regarding lift are certainly appropriate in database marketing. Achieving ever-greater efficiencies is critical to producing the largest return on investment (ROI). Usually, the argument is to shave small costs and mail to break even or target best prospects to maximize profit.
One of the best ways to illustrate lift is through a graphic called a lift or cumulative gains chart. Along the horizontal axis, one plots the percentage of the population to be contacted. Along the vertical axis, the cumulative percentage of positive results expected is plotted. With a randomly drawn campaign, one would expect to get 50 percent of the possible positive responses by contacting 50 percent of the population.
Lift charts are normally used for predictive models when the criterion is a continuous value such as a propensity score or lifetime value (LTV). Although lift charts are not usually used to measure the effectiveness of targeting specific segments, they certainly can and should be used for this purpose.
In this application, segments are groups of individuals in the same category. Membership in these segments is only nominal. This use is very different from a predictive model where each individual has a singular score. Consequently, to produce a lift chart, one rank orders the segments, rather than the individuals. Some segments are better targets than others because they offer higher profitability or have some particular need the marketer can easily satisfy.
To illustrate these relationships, a hypothetical marketer of consumer products on the Internet can divide his market into some number of segments according to a set of criteria that includes attitudes, behaviors and demographics. Let's say that he has three segments ranked in order of their expected payoffs. Segment A is 34 percent of the market, but accounts for 52 percent of the expected buyers. Segment B, on the other hand, is 29 percent of the market and 27 percent of the expected buyers. Segment C, the largest segment, is 37 percent of the market, but only represents 21 percent of the buyers.
The lift chart in Figure 1 shows the effectiveness of targeting segment A first and then segment B. To achieve 52 percent of sales and expend resources contacting 34 percent of this marketer's database certainly maximizes marketing efficiency.
Figure 1: Lift Chart
In this application, segments may be derived by any number of means such as CART and CHAID trees to determine best prospects. This technique uses a criterion score to split a population, finding nodes or groups of attributes that distinguish those with high criterion scores from those with low scores. By ranking the nodes from best to worst as if they were segments, one can graphically show the lift expected from the analysis.
Recency, frequency and monetary value (RFM) are frequently used to evaluate prospects in a campaign. Segments based on RFM can also be measured according to a lift chart. The sample is rank-ordered by each of the measures and divided into quartiles or some other percentage breakdown. Recent buyers are the most likely to respond to additional or further promotions, frequent buyers respond at a higher level than less frequent buyers and big spenders respond better than small spenders. By characterizing these RFM segments by expected lift, management has a better idea of the effectiveness of a campaign.
As one would calculate with a traditional lift chart using a predictive measure, one can use the chart to determine return on investment. In the case of the Internet consumer products marketer, his "in the mail" costs are $1 for each of 100,000 contacts, with an expected purchase rate of 3 percent. To break even, he must realize $33.33 in profitability on each customer. However, Segment A accounts for 33 percent of the market and 54 percent of sales. Therefore, Segment A is likely to respond at five percent or 64 percent higher than the average of three percent. Consequently, targeting this segment produces a breakeven of only $20.
Having developed segments and predicted their value and expected ROI on a lift chart, the next step is to measure the results. Quite often, modelers predict the results but fail to follow up with subsequent results. It is important to "close the loop." One should determine the actual results of the campaign, which segments produced the best results and the actual return on investment.
Where's the lift? Segmentation, when performed correctly, can produce a considerable return on investment. By selecting the correct targets, tailoring the most effective message and marketing products and services with high appeals to the targets, lift is almost a certainty.
Segmentation can greatly improve an enterprise's marketing efficiency. By using techniques such as cumulative gains or lift charts, modelers can demonstrate graphically to their management the value of their models. Doing so also imposes the discipline of measuring the impact of using segmentation. Only through such tangible analysis will management realize the size of positive impact of this approach on the bottom line.
Register or login for access to this item and much more
All Information Management content is archived after seven days.
Community members receive:
- All recent and archived articles
- Conference offers and updates
- A full menu of enewsletter options
- Web seminars, white papers, ebooks
Already have an account? Log In
Don't have an account? Register for Free Unlimited Access