Particle.news

DraftKings Used Machine Learning to Target High‑Value Losing Customers, Reports Say

The disclosures raise questions about the company’s decision to abandon internal tools meant to predict gambling crises, a choice that may draw regulatory scrutiny.

Overview

  • The New York Times investigation published Monday reported that DraftKings built a 2023 machine‑learning model that scored users on “elasticity,” a measure of how likely a customer was to respond to promotions.
  • Former employees told reporters the elasticity score steered bigger offers to customers judged most likely to generate net revenue, including players who went on to lose large sums.
  • Separate teams developed predictive models in 2024–2025 intended to flag customers headed for gambling crises, but presentation meetings were canceled and the projects were shelved, according to former staff.
  • DraftKings has denied improper targeting and said it evaluated risk models and chose not to deploy them because they were not evidence‑based while pointing to existing responsible‑gaming tools.
  • The reporting notes the business scale behind the practices—about $8.7 billion in gross revenue and roughly $3 billion in promotions in the cited period—and highlights personal harm, such as a gambler who lost nearly $70,000 and said repeated offers undermined his recovery.