How Favor Delivery transformed mobile observability with bitdrift

Prior constraints

Before using bitdrift, the Favor team was limited to 5-10 minutes of breadcrumb logs at best, and just seconds of logs at worst.

Senior Staff Software Engineer, Igor Korobka, explains, “That wasn't nearly enough to understand some of the more complex problems. It would help for some of the more obvious ones - except when it cut off the head of those logs - but in many cases the information was like getting a couple of grains of sand to explain the whole sea.”

Getting logs for non-fatal errors

One of the biggest challenges the Favor team faced was not with crashes at all, but rather unknown issues that would impact only some users. Igor shares, “we had an application that was not crashing, but people would still report problems. And [without a crash], you don't have the only point of reference that was giving you logs. No crash, no logs. We essentially were hitting a wall.”

With bitdrift, that changed. Suddenly, the team could capture vast amounts of mobile telemetry data. The analytics and increased visibility that came with bitdrift has shifted how the team works. “bitdrift has had a very transformational change for the entire company” says Igor, “we started with essentially all the mobile telemetry analytics we had coming from crash reports. Now, we can answer some very quantitative questions that were not possible before like, what's the baseline number of users you get on this specific screen? What happens if it deviates up or down? When you add this feature to the screen, does that number change at all?”

Principal iOS Software Engineer, Alex Koshy, adds, “we had this big mountain of non-fatals all lumped under a single crash entry that no one really paid attention to. Now, with bitdrift, we have very distinct workflows tracking isolated problems.”

Unlocking this level of real user data beyond the bounds of a crash - with the ability to trigger collection ad hoc and without an app release - has meant that the Favor team has more insight into exactly what their users are experiencing. On top of that, they can proactively solve issues before a crash happens, and understand & fix issues in the UX that could otherwise create negative experiences for their customers.

Uncovering & facing down skeletons

Beyond the issues the team knows to investigate, bitdrift has helped uncover issues the team didn't know they had. Igor shares, "the first time bitdrift really clicks for anyone is when they discover something they didn't know existed. I came across this multiple times, like, for example, when we saw excessive application process terminations in the background that affected a lot of our metrics, or when we saw different latencies for seemingly the same requests between iOS and Android, or different error rates on different network endpoints."

Igor shares, "In many cases, what we learned after we started using bitdrift, is that we didn't realize the problem existed in the first place." This increased insight means the team can create better, more reliable in-app experiences for their users at a level that many other teams fail to reach.

Validating the user experience in the wild

One of the biggest wins for the Favor team has been the ability to confirm that changes they make to the product are working in the wild. Staff Software Engineer, Daniel Kaparunakis, says “I've enjoyed the ability to validate the work that I'm doing. It's great to attach a workflow to a new change and be able to say ‘look, the numbers support that the change is working as intended.’” This is enabled by bitdrift’s ability to rapidly propagate workflows into the app population and collect data, which enables users to get information back near real-time.

When things are working properly, the team can use bitdrift as an extra layer of confirmation. And, when things aren’t working as expected, bitdrift shifts into a debugging tool.

Alex Koshy, shares “it’s really easy to think back on many incidents where, sure, we were seeing a screenshot from a user that shows the modal, but the best we could do was give some context on code. Now, with bitdrift, we can be an active participant in incidents, which I'd say is pretty revolutionary.”

Daniel shares, “there's a whole vertical of things that we're finding where we strongly believe something is happening… okay, now prove it. With bitdrift, we can know what’s happening without a shadow of a doubt.”

Having bitdrift in their toolkit has increased confidence around new releases. “Having a tool that enables you to actually take steps when things don't perform as you expect is huge” says Daniel, “the bottom line is, as leaders of the mobile space at Favor, we can now look at our executives and say ‘I know that X, Y, and Z is going to perform this way because here's my observability stack telling me so,’ that's really powerful.” And the team can get that validation without adding analytics events to the app, and without needing to re-release the app.

A user-friendly interface

The Favor team has been particularly successful in rolling bitdrift out to their larger team.

Daniel Kaparunakis shares, “using workflows is significantly easier than having to learn a very specific query language that works on one specific tool. I can easily tell my engineers, ‘hey, here's how you use the tool.’ Then boom.”

Daniel, Igor, and Alex even hosted a live workshop to train their teams on how to use bitdrift, where they walked through building workflows to uncover and debug issues. Daniel shares, “we were able to go from zero to value in an hour.”

Conclusion

Not only has bitdrift enabled the Favor team to get significantly more insight when a crash does happen, but perhaps more importantly, they’ve been able to uncover previously unknown issues, speed up debugging with more context, ask deeper questions around product performance, validate the stability & effectiveness of new releases, and ultimately create a superior mobile user experience.

Igor shares, “I think bitdrift is one of those rare products where people have needed it desperately, they just didn't know it. It makes a previously very expensive and hard to obtain context available to almost anyone, which will propel the entire industry to new heights and new levels of quality.”