Lyft Holiday Tipping: A Lifecycle Messaging Experiment
Measuring a post-ride message against a no-message control
A two-week holiday experiment from Anthony Sledge’s work at Lyft tested post-ride messages with riders who rarely or never tipped. The retained account reports about $80,000 in additional driver tips across 23,000 additional tipped rides during the test.
At a glance
~$80k
Additional driver tips
Observed two-week test, not annual revenue
~23k
Additional tipped rides
Reported experiment effect
2 weeks
Test window
Holiday period in late 2024
Capabilities used
The problem
The team wanted to improve tipping among eligible riders who rarely or never tipped, while learning whether a seasonal message benefited from added social proof.
The approach
The work and the decision behind it
The experiment compared a feel-good seasonal message, the seasonal message with regional social proof and a no-message control. Timing and audience eligibility were defined before evaluating incremental tipping.
Results and limitations
The reporting window and what it supports
Across the two-week test, the retained account reports approximately $80,000 in additional tips and 23,000 additional tipped rides. The simpler seasonal treatment performed better than the social-proof version in the reported comparisons. A $3.9 million annualized projection appears in the career account; it is not observed annual performance, and the two-week result should not be assumed to persist year-round.
Why this matters
The relevant decision is whether a specific message helps a defined audience at a useful moment. This case supports measuring incremental behavior and testing additions such as social proof rather than assuming they improve every message.