Uplift Modelling
A modelling technique that predicts the incremental effect of a marketing action, so effort targets players whose behaviour it will actually change.
Definition
Standard propensity models predict who is likely to act; uplift models predict who will act because of the intervention. By comparing treated and untreated (control) groups, they classify players into 'persuadables' who respond only if contacted, 'sure things' who would act anyway, 'lost causes', and 'sleeping dogs' who react negatively to contact. Targeting only persuadables avoids wasting bonus spend on people who would have deposited regardless and avoids annoying or harming those best left alone. It is a sophisticated CRM approach that improves campaign ROI and, used responsibly, can reduce unnecessary contact — including with players who should not be nudged at all.
Worked example
An uplift model might reveal that a bonus offer only changes behaviour for 12% of a list; sending it to just those persuadables delivers most of the incremental deposits at a fraction of the bonus cost.
Why it matters
For a learner, it introduces the crucial difference between correlation and causal incrementality. For a professional, it is the state of the art for spending marketing budget only where it changes outcomes, with the side benefit of not contacting 'sleeping dogs'.