RESOURCES / SOLUTION SPOTLIGHT

Traditional Response Models

There is a subtle but important difference between

Targeting people who would buy if they are part of a campaign
&
Targeting people who would only buy if they are part of a campaign

Conventional Response models focus on the former and identify customers who are likely to purchase a product if targeted in a campaign. However the model does not identify customers who would have purchased even if they were not included in the campaign. As a result, conventional response models spend precious dollars contacting customers who would have purchased eitherways (Sure Things).

Uplift Modeling

In the latter approach, we would focus on customers who would purchase only if they are targeted through a campaign. This radical type of modeling is referred to as Uplift modeling.

Measuring Uplift

For any campaign, we would see that the population could be broadly divided into four segments

Persuadable: Segment on whom campaign will be most effective, meaning that their probability of purchase will increase if they are included in the campaign

Immovables: There will others for whom the campaign has little or no impact which can be further subdivided into

Sure Things: Customers would purchase whether they are part of campaign or not

Lost Causes: These customers would not purchase even if they are targeted.

Sleeping Dogs: There could also be a segment for whom the campaign will have a negative effect, i.e. they will be less likely to purchase the product in question as a result of the intervention

Evidently, marketing efforts should focus on the Persuadables and avoid wasting effort on the other segments. If we look at incremental sales generated by targeting the segments in the order of Persuadables, Immovables, Sleeping Dogs, we can see that initially incremental sales would improve, then stagnate and then actually fall (as depicted below).


Test-Control Methodology to Predict Uplift

For uplift modeling, ideally we should have similar looking customer segments, who have been targeted as well as not, and the responses could be used to identify the effect of campaign on these segments. To develop uplift models, two representative samples are drawn from the population to form the Test (targeted by the campaign) and Control group (not targeted by the campaign).

From the positive respondents for the campaign on the Test group, we will develop a predictive model to predict the propensity of purchase, which will be denoted by

pT = probability ( purchase | Campaign)

Another predictive model would be developed on the control group population to determine the propensity of customers to purchase and this is denoted as

pU = probability ( purchase | no Campaign)

Based on the ouput for these models, we can determine the uplift in propensity for any customer if he is included in a campaign. Based on the magnitude of incremental lift in purchase propensity, customers could be targeted for the campaign.

Benefits of Uplift Models

LatentView has seen signficant returns by adopting the Uplift modeling techniques to develop response models vis-à-vis conventional response models.

For one of our engagements, LatentView developed and compared the performance of the traditional response models with uplift models and we have seen that the campaign profitability dramatically improves by atleast 50% (in the top 2 population deciles) as shown in the figure below.

Therefore, by using Uplift modeling, we can gain significantly by targeting a smaller population,thereby improving campaign performance!