How Lyft reduced mobile MTTR by over 80% with bitdrift

• Slashed the time it takes to find and fix complex bugs from several weeks to a matter of days

• Comprehensive, log-everything telemetry helps engineers quickly track down the data they need as soon as they need it

• Single, unified view provides comprehensive awareness, enabling rapid and intuitive analysis of issues

A leading rideshare company, Lyft (a bitdrift investor) provided access to 709 million rides to more than 40 million riders in 2023. The company’s mobile app offers access to rideshares, bikes, and scooters as well as onboarding and routing for Lyft’s millions of drivers – its complexity and its large user base serve as an excellent example of the types of challenges bitdrift was built specifically to address.

Transforming day-to-day operations with comprehensive telemetry

With bitdrift, teams can log everything (and we mean everything) without worrying about prohibitive infrastructure bills. Unlimited telemetry completely transformed the way Lyft’s mobile payments team debugged and handled issues: the relevant context for a crash is available instantly, without needing to deploy changes to collect more data.

Artem Havriushov, an engineer on the Lyft payments team, is able to resolve issues significantly quicker with all of the telemetry he needs at his fingertips. “Since we’re precisely logging everything, I can recreate the user flow to understand which part of the flow is broken,” he says.

Massive volumes of telemetry doesn’t mean that bitdrift is hard to use; actually, quite the opposite.. While Lyft has competing tools available internally, most of them require the assistance of data scientists – bitdrift doesn’t. “It’s super simple, and the UI is intuitive,” he says. “When I first wanted to build a workflow, I started to read the documentation – but then I realized I didn’t need it.”

Session Replay quickly pinpoints complex issues

bitdrift’s Session Replay gives developers a privacy-first, high fidelity, storage-conscious wireframe recreation of individual mobile screens. For a mobile app as complex as Lyft’s, bitdrift’s Session Replay functionality has been particularly useful in tracking down the specific source of an issue. “I really love that you can replicate views, screens – everything the user sees while they go through their own flow,” Havriushov says.

In one case, he recalls, he needed to determine precisely what screen the user was on when a bug was being triggered, but the app’s complex structure made it hard to figure out exactly what was being shown.

“Using Session Replay, I caught the precise view – and I found the issue,” he says.

The fidelity that bitdrift’s Session Replay gives Lyft helps them rapidly uncover issues that would otherwise have gone by unnoticed or taken days to resolve.

Using Funnels to squash bugs in days, not weeks

Recently, as Lyft’s developers worked to add significant new functionality to the onboarding process for drivers, there was a clear difference in a/b testing results between the iOS and Android versions of the app. “We realized that we had missed something and that one of the platforms was working incorrectly,” Havriushov says.

With the new functionality already live, they urgently turned to data scientists to track down the issue, but the length and complexity of the user flow made that process much too time-consuming. “We began to test everything from start to finish, but it’s very hard to do so, because the flow can create different branches,” he says.

Havriushov then turned to bitdrift, and solved the problem in a couple of days.

“I created two workflows for iOS and Android,” he says. “I replicated a variety of flows during onboarding, and I created two funnels to check them.”

The Funnels feature quickly pinpointed the issue: if Android users left the onboarding process and returned to it, they were sent back to the beginning of the flow rather than to the point where they’d stopped, whereas iOS users were able to pick up where they’d left off. The result was a significant dropoff in signups on Android compared to iOS – and the rapid resolution of the issue instantly improved the user experience for drivers on Android.

Without bitdrift, Havriushov says resolving that issue would have taken much, much longer. “To build a similar flow using other tools would take up to a week, and then to debug that flow and understand the data would take a week to a week and a half,” he says. “With bitdrift, I could do it in days: one day to set up the workflow, then wait a day for the data to be collected, then create an analysis.”

The experience was so positive that Havriushov’s team is now doing the same thing preventatively every time new functionality is introduced. “As a safety measure, whenever we build a release, we build two flows for iOS and Android platforms just to check that the rollout goes smoothly,” he says.

In every way, he says, bitdrift’s functionality has been transformative. “It’s a game changer. It’s simple, understandable – and it works.”