← Back to the feed
Chips & computingResearch findings

A Stopped Random Walk Model for Long-Term Treatment Effect Estimation in Advertising

Original published 2026.02.06 03:43 KSTAdded to Blackdeck 2026.10.21 23:29 KST

What has changed?

A stopped random walk model is introduced to account for user behavior and advertiser bidding policy changes in advertising experiments

Key points

  1. A stopped random walk model is proposed to estimate long-term treatment effects in online advertising systems
  2. User interaction trajectories and advertiser bidding policies are explicitly modeled as dynamic responses to treatment
  3. Confidence intervals for long-term effects are constructed using budget-splitting design and central limit theorem

Connect it to your work

The model considers long-term trajectory shifts in user interactions and advertiser budgets

What remains unproven

This research note is based on an abstract; check the original for test conditions and application scope.

Originals used for this note

Links open the original publisher in a new tab.

  1. arxivCausal Inference on Stopped Random Walks in Online Advertising