How ThredUp resolves 60% of mobile issues before a human ever sees them

The problem: 80% of troubleshooting time spent trying to reproduce issues

ThredUp is one of the largest online consignment and thrift platforms in the world, transacting more than 6 million orders a year across 55,000+ brands and 1.65 million active buyers. The mobile app experience is critical: it’s responsible for roughly half of all revenue for the global online consignment and thrift store. In 2025, that put more than $150 million in revenue in the hands of a 7-person mobile engineering team.

This leaves little slack for chasing bugs that can't be reproduced; a problem that came to a head in 2025. ThredUp’s mobile engineering team allocates roughly 80% of their time toward building a better customer experience, and 20% to solving customer issues. They’d been using Firebase to flag app crashes, but had a harder time understanding why a shopper's cart mysteriously failed at checkout or why a listing hung mid-load.

"If we couldn't reproduce it, we didn't have anything to go on," explained Valerii Kuznietsov, who leads mobile engineering at ThredUp. “We’d spend 80% of our troubleshooting time just trying to reproduce the issue.”

The team needed a tool that gave them full transparency into what was actually happening on-device, without adding process friction or scaling cost linearly with data volume, which is where their other observability tools started to break down.

Full resolution mobile observability with bitdrift

When the ThredUp team kicked off a formal initiative for mobile observability, they evaluated 8 vendor solutions against 3 main decision criteria:

• Full data transparency
• A cost model that would make sense over time and at scale
• Minimal runtime performance impact

That last point was a big one, and one of the main reasons they went with bitdrift.

“We already had customers reporting performance issues, so we didn’t want a mobile observability solution that would add more performance issues,” said Kuznietsov. “Except for bitdrift, all the others had a performance impact on our application, ranging from minor to significant. It’s because they try to be everything and focus on all areas of mobile, including making video recordings of user experiences, and that causes significant performance impact. bitdrift didn’t impact our application performance at all and fit perfectly for the slice of functionality we were looking for.”

Another benefit of bitdrift was how naturally the mobile observability platform slotted into existing workflows and technologies, for example, via GraphQL support. About 90% of ThredUp's mobile requests run through GraphQL; bitdrift's native, first-class support for GraphQL, with the ability to break out networking logs by operation name, stood out.

ThredUp got fully instrumented and into production during their proof of concept with bitdrift. The team finally had one clear picture of what was happening on any device, tied to the actual session instead of an aggregated crash count. Engineers can pull the exact session behind an issue and see the full sequence of events that led to it, on demand, instead of waiting for a user to describe what went wrong.

ThredUp is also using bitdrift's entities feature to zero in on its daily-active power users. Those are the shoppers who use the app constantly and are most exposed to the subtle logic and UI issues that standard crash reporting tools were never built to catch.

Now, AI agents resolve issues before they reach Support

ThredUp was an early adopter of bitdrift’s agentic investigation capabilities, using the bitdrift CLI and public API to allow AI agents to iteratively query and analyze telemetry from all customer devices. bitdrift AI allows ThredUp to filter out the data that's just noise, which speeds up resolution and keeps processing costs down.

The team appreciated, once again, how easily bitdrift integrated with their existing AI pipelines, letting engineers query and interact with data using plain language. "It functioned like a natural extension of what we were already doing, not another tool we had to work around," said Kuznietsov.

Here's how it works in practice:

1. ThredUp tracks business-critical events in bitdrift, like add-to-cart actions and checkout failures.
2. When one of those flows throws an error the team hasn't seen before, an AI agent starts working the problem immediately, pulling user flow data from ThredUp's internal systems to piece together what happened.
3. In many cases, the issue is understood and on its way to resolution before a customer support agent has even joined the ticket.

The result: 60% of technical mobile issues are now resolved by AI agents with no human intervention at all.

“With bitdrift, we’ve brought infinite scale to problem resolution. Because we have enough data and AI to help us, we can help 90-95% of customers experiencing minor problems instead of the 20-30% that we could support before,” said Kuznietsov.

Previously, the team burned most of its support time just trying to reproduce a problem. Now the data is already there.

The takeaway

bitdrift gives ThredUp's small mobile engineering team full visibility into their apps. AI agents now close 60% of customer issues without a person involved, and the team’s support workflow is 5X more efficient.