An Experimental Study on Platform Steering

How to cite

Wang, Yifei; Tucker, Catherine; Hao, Yulin; Luo, Yan. 2026. "An Experimental Study on Platform Steering." AEA RCT Registry. AEARCTR-0018200. https://doi.org/10.1257/rct.18200-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Yifei Wang* , Catherine Tucker , Yulin Hao , Yan Luo

* Primary investigator

Registry ID
AEARCTR-0018200
DOI
https://doi.org/10.1257/rct.18200-1.0
First registered
Last updated
Status
In development
Trial dates
2026-08-12 – 2026-10-05
Countries
Private
JEL codes
L11, L13, M31
Registry record
http://www.socialscienceregistry.org/trials/18200

Abstract

The extent to which platforms have power over participants has been a central concern in recent competition policy debates. One key channel through which platforms can exercise such power is platform steering — the ability to shape buyers’ choices among competing sellers on the platform. In this paper, we study the steering power of a digital aggregator.

We partner with a leading bus ticketing aggregator platform in Bangladesh to conduct a large-scale field experiment. Users are randomly assigned to one of four conditions. In the company-specific discount condition, users receive platform-sponsored discounts for tickets for promoted bus companies. In the messaging treatment, users are shown promotional messages featuring specific bus companies, but without any price discounts. In the generic discount condition, users receive a generic platform-sponsored discount that does not reference any specific company. In the generic message condition, users receive a generic promotional message that does not reference any specific company.

We examine the extent to which these platform interventions steer users toward the promoted sellers. We also investigate how the strength of this effect varies with the type of steering (discounting vs. promotional message), user loyalty to the platform, loyalty to individual bus companies, the level of geographic competition in the local bus market, as well as how platform promotions interact with the platform ranking system.

Intervention

The intervention involves sending SMS messages to consumers in the country who have previously booked bus tickets through the platform at least once since 2025. Depending on the treatment condition to which a user is assigned, the SMS either includes a coupon code that waives the platform usage fee or contains a generic promotional message.

Intervention #1
The platform sends users an SMS message that includes a coupon code which, when applied, waives the platform fee for any ticket purchase on the user’s most frequently traveled route.

Intervention #2
The platform sends users an SMS message that includes a coupon code which, when applied, waives the platform fee for ticket purchases with a specific bus company on the user’s most frequently traveled route. The name of the bus company is explicitly mentioned in the message.

Intervention #3
The platform sends users an SMS message encouraging them to book tickets on the platform with a specific bus company when traveling on the user’s most frequently traveled route. The name of the bus company is explicitly mentioned in the message. No coupon code or other discount is provided.

Intervention #4
The platform sends users a generic SMS message encouraging them to book tickets on the platform when traveling on the user’s most frequently traveled route. No coupon code or other discount is provided.

Experimental design

We evaluate whether and how different types of platform promotions affect platform usage and users' product choices. Users are randomly assigned to one of four conditions based on their historically most frequently traveled bus route, as identified from their past bookings on the platform. In the company-specific discount condition, users receive platform-sponsored discounts for tickets for promoted bus companies. In the company-specific messaging condition, users are shown promotional messages featuring specific bus companies, but without any price discounts. In the generic discount condition, users receive a generic platform-sponsored discount that does not reference any specific company. In the generic message condition, users receive a generic promotional message that does not reference any specific company.

Design details

Randomization method
Randomization done in office by a computer
Randomization unit
Bus route, defined as an unordered pair of cities.
Clusters
88 routes (city pairs)
Observations
We plan to send SMS messages to approximately 2009 users per city pair, totally 176,790 users. However, some SMS delivery failures are expected. The final sample will therefore consist of users for whom the platform successfully delivers the SMS message.
Arms
22 routes (city pairs) for each of the four conditions.
Primary outcomes
Platform usage: the proportion of users who purchase bus tickets via the platform, among all users who purchase bus tickets, within a certain period Company choice: the proportion of users who purchase the promoted bus companies, including in on- and off-platform purchases, within a certain period Site visits: the number of pageviews to the website and app of the platform per route, the number of active users who visit the website and app of the platform per route, the number of pageviews and the average enagement time among users who visit the website and app of the platform per route, within a certain period
Secondary outcomes
Coupon redemption: the number and proportion of (incremental) purchases that applied the coupon code among purchases from customers who have access to the coupon code. This measure is used to distinguish between two potential mechanisms underlying the effects of platform price promotions. A high redemption rate would indicate that users actively respond to price reductions, consistent with the platform having pricing power. In contrast, if treatment effects on platform usage or product choice occur primarily without coupon redemption, this would suggest that promotional messages operate mainly by attracting user attention rather than through realized price discounts.

Cited by

No citation data available for this registration yet.

Keywords