Some frequent Lyft customers declined subscription offers that would have saved them money, while some subscribers did not use the service enough to break even, researchers found in experiments on subscription pricing.
The National Bureau of Economic Research working paper examines a gap between the plan that appears best after a period of use and the one a customer chooses beforehand. It was revised in September 2026 and has not been presented as a universal result for subscription businesses.
Aaron Bodoh-Creed, Brent Hickman, John List, Ian Muir and Gregory Sun use Lyft experiments and an economic model to study how customers choose between a subscription and paying without one.
Future use is hard to predict
A subscription typically involves a fee paid in return for access or lower charges during a set period. Whether it saves a customer money may depend on how frequently the service is used and what each purchase would otherwise cost.
For a simple illustration, a plan costing $10 that saves $2 per trip needs more than five qualifying trips to produce a net saving. A person making fewer trips would pay more under that arrangement, assuming no other benefits or fees.
Customers make the decision before they know exactly how many trips they will take. The researchers model several possible influences, including mistaken expectations about future demand and the prominence of the subscription offer.
The results do not establish why any particular individual chose a plan. A customer may value flexibility, dislike an upfront payment or have information about future needs that a company cannot observe.
What a model says about pricing
In the authors’ modeled comparison, a subscription program without the estimated choice frictions would increase profit by 5.72% over an optimized per-use pricing approach. Once the researchers included those frictions, much of the modeled gain disappeared.
That 5.72% is a counterfactual model estimate, not Lyft’s reported profit increase or an expected return for another company. The result depends on the model, the experiments and the market studied.
The paper discloses that List was Lyft’s chief economist and Muir and Sun were Lyft employees when the experimental research was conducted. The authors report that none currently work for Lyft.
For a business designing a plan, the work points to a practical measurement problem. A sign-up rate alone cannot show whether customers who would benefit understand the offer or whether subscribers are receiving the value they expected.
Companies can compare the plan fee with actual qualifying usage, track support requests and cancellations, and explain clearly when charges begin and how to cancel. Testing changes against a comparable group can help separate the effect of an offer from differences among the customers who choose it.
Our earlier coverage of subscription activation explored when customers receive access. The Lyft research examines a different moment, the choice of pricing plan, and why a customer may find it difficult to anticipate future use.
