Parallel Experimentation and Competitive Interference on Online Advertising Platforms

For your marketing

Possible founder application

A useful caution for campaign tests run in competitive auctions: your measured lift may depend on what other advertisers are doing simultaneously. Coordinate or repeat tests when market conditions shift.

Reported result

What the study found

Ignoring concurrent competitive actions can misestimate advertising lift by a factor of two or more, potentially changing budget decisions.

Research question

What the authors investigated

How can advertisers measure ad effects when multiple firms are experimenting at the same time on a platform?

Design

Develops an approach that accounts for parallel advertiser experiments and evaluates it using an experimental design on JD.com.

Sample and setting

Advertising experiments on JD.com, a large e-commerce platform and digital-ad publisher; exact sample counts are not in the abstract.

Scope and transfer

Limitations to keep in view

The publisher abstract does not state limitations or sample counts. The setting involves a major marketplace with overlapping advertiser campaigns.

Research sources

INFORMS publisher abstract retrieved with Firecrawl; the reported marketplace setting and effect-size warning are abstract-level.