Optimizing Onboarding to Drive User Engagement
How Levered helped Lyft's growth team run a controlled experiment to eliminate driver confusion and optimize the onboarding experience through data-driven communication strategy.
The problem
• New Lyft drivers received a Day 6 earnings summary email early in their lifecycle — before many had enough context to interpret the numbers correctly. • The growth team suspected it was creating confusion rather than clarity, but had no evidence to act on. • ~16,000 drivers per group were in scope — changing lifecycle comms without a controlled test risked core health metrics at scale. • The org needed a defensible answer, not an opinion, before touching a high-volume driver communication.
The fulcrum
The highest-leverage change we could make
• Designed a 4-week holdout: treatment group kept the email, control group did not receive it. • Monitored two signal types simultaneously — confusion signals (help page views, support tickets) and hard engagement metrics (driver hours, rides, earnings). • Drafted parallel decision plans before results came in: if holdout won, strip the email and move payout education in-app; if not, refine copy and targeting instead of guessing. • Structured the experiment so the result was actionable regardless of which direction it went.
Applying force
What happened, and what we scaled into
• Produced a statistically defensible answer on whether the Day 6 email helped or hurt early driver activation. • Gave the team a clear path: remove inbox noise, shift education to the right surface (in-app), and keep early-driver momentum intact. • Decision made with data, not opinion — protecting thousands of drivers per cohort from a change that hadn't been validated. • (Detailed lift figures remain Lyft-internal per confidentiality.)
Why this matters
• At Lyft's scale, a gut-feel change to a lifecycle email touches tens of thousands of drivers per week. The cost of being wrong is measured in driver hours and support volume. • The value here wasn't running an experiment — it was designing it correctly: right metrics, right holdout structure, right decision framework before the data came back. • Most growth teams can run A/B tests. Fewer can design experiments where the result is actionable no matter what it shows. • If your lifecycle comms are based on assumptions rather than evidence, this is the kind of work that changes that.
Most engagements fail at the handoff between strategy and execution. Levered doesn't hand off — we own both.