Uplift Modeling: Making Predictive Models Actionable

Mike Thurber

April 14, 2017


BLOG_Uplift-Modeling-Making-Predictive-Models-Actionable.jpgPredictive models typically estimate the likelihood of future events, such as whether it will rain tomorrow or which customers are most likely to “churn” by cancelling their phone contract.  In the case of the weather, we do not expect to change it; we just want to know how to adapt.  However, the goal for most use cases is to be more proactive; we want to understand what action to take to change the outcome in a favorable way. In these cases prescriptive, not just predictive, analytics is required.  The return on investment comes directly from knowing the impact of alternative treatments. By knowing the impact of each treatment, resources can be targeted where they will be most effective and withheld where they will have negligible effect or worse, have a negative effect.  This great objective of data science, to intelligently drive day-to-day business decisions based on data, is the purview of uplift modeling.  This blog will explain what uplift modeling is and why it can be much better than directly modeling the outcome.

There are many applications of predictive modeling where the outcome is predicted as advice only to a human decision maker, and no action is directly taken automatically from the model result.  An example is workload prioritization for customer service or fraud investigative teams. But where we can, we aim to influence the outcome one way or another. Will a live agent offering the phone customer a contract upgrade decrease their likelihood to churn?  Will soliciting a fund raising prospect with a flyer in the mail improve their chances of making a donation?  Will offering a moving bonus increase the likelihood that a desirable candidate will accept our employment offer?

Uplift Modeling Examples.png

Uplift Modeling is Not Directly Measurable

Uplift modeling is also known as incremental modeling, treatment effects modeling, true lift modeling, or net modeling.  Uplift is the increase in likelihood of the outcome with the treatment as compared to the outcome without the treatment.  We can’t observe this difference, or causal effect, directly, but must infer it from an experiment. It is very helpful to visualize a 2x2 matrix, as shown below, with four categories of people (say) to be classified, as: (a) Persuadable, (b) Sure Thing, (c) Do-Not-Disturb, and (d) Lost Cause as shown in the figure below.

Uplift Modeling 2x2 Classification Matrix.png

Uplift modeling’s objective is to find Persuadables. Then, you can target resources on the cases that are likely to be positively impacted by the treatment.

Uplift Modeling Benefits

Uplift modeling can apply to any modeled outcome, human or not, such as fertilizer on crop yield, a drug on a patient’s health, retail loyalty programs on profit, or messages in political campaigns.  Whenever treatment resources are limited or there is a possibility of a negative treatment effect, uplift analysis is an effective tool. Uplift analysis models the effect of treatment, rather than the outcome directly. If we know how likely something is already, and how likely we are going to be able to change it with a treatment, we can classify prospects as either “sure things”, “persuadables”, “lost causes”, or “do not disturbs”.  This is extremely valuable as a way to get the most out of ones analytics investment.

Request a consultation to speak to a data analytics consultant about how Elder Research can help drive better insight from your analytics projects.

Previously published on Predictive Analytics Times.


If you would like to learn more about the benefits of uplift modeling download the white paper.

About the Author

Mike Thurber Lead Data Scientist Mike Thurber is an expert data analyst, comfortable with diverse data sources in diverse industries, and extracting relevant and valuable insights from available data. His modeling work ranges from predicting high payouts on long term care claims to identifying healthcare provider fraud to measuring the effect of Cesarean delivery on infant health. His broad experience managing a variety of analytic initiatives consistently generates business value through expert collaboration, data integration, insightful data analysis, statistical testing, and predictive modeling. Other examples include gleaning insights on how complex consumer choices impact sales, predicting profitability of prospective customers, calculating fraud and financial risk of many kinds, showing how call center interactions affect customer retention, forecasting recovery of losses due to loan default, modeling maintenance events on natural gas wells, and predicting propensity to make voluntary monetary donations. Mike earned a BS degree in Chemical Engineering from Brigham Young University and a Master's degree in Statistics from Virginia Commonwealth University.