Lessons from Growth: Small Data, Big Insights

We recently gave our onboarding flow a facelift. This is something you may not have noticed unless you set up a new Sentry account every other week since onboarding is simply the process of signing-up, creating a new project, and sending data to it for the first time. We made cosmetic upgrades for a cleaner look, plus began collecting one extra piece of data that we’ve never requested at sign-up before:

Platform. Are you coding in JavaScript? Ruby? Or perhaps the always popular Other?

Choose a language or framework
Pick a platform, any platform

The biggest step in getting Sentry up and running is integrating it with your platform, so asking this one small question has led to big insights about the ease of using different platforms with our service.

Leading platforms: Then and Now

Before we dive into these new insights, let’s get some context on what our platforms look like right now.

While Sentry started off as a way of debugging errors in Django (aptly named django-db-log), we’ve since evolved to support 41 languages and integrations. What was popular 5 years ago looks very different now as our user base has grown and the industry itself has evolved. The single largest trend we’ve seen in the industry is the increase in javascript usage and it’s now our #1 language (accounting for close to a third of all Sentry projects).

Projects per platform
Popular Sentry platforms — the larger / darker the bubble, the more projects on that platform

Onboarding variation by platform

Our biggest goal is to ensure each user successfully integrates Sentry regardless of which language they choose for their project. One of the key ways we measure success is by looking at metrics on whether a project has sent a test or real event (First Event). We track how many users send an event, as well as how long it takes them to do so.

So how are we doing? Thanks to the new platform data we are collecting during onboarding, we can dig into variations in First Event success across platforms. Let’s first look at the number of users that send their first event under each platform (and by numbers we mean bubbles because…proprietary data). Each bubble represents the share of users within that particular platform; the bigger the bubble and the darker the color, the higher the success rate. Python and PHP projects come out on top in this case, with roughly 70% of users on these platforms making it through to First Event within two days of sign-up.

First event success
First event by platform — larger, darker bubbles have a bigger share of successful projects

We can take this data a bit further and consider how long it takes for users of different platforms to reach a First Event. PHP is far ahead of the pack here, with users taking about 90 minutes to reach First Event on average Python, JavaScript and Ruby also perform well, with users sending their First Event within two hours. Of course, in all of these cases, this means that a lot of users are getting things moving in less than an hour, followed by a long tail of other users sending their First Event at some point over the first week.

Time to first event
Time (mins) to first event — darker colors represent a larger number of successful projects

Conclusion

Sentry supports many languages and integrations, and this new data will give our engineers insight in how we can make improvements that simplify how users integrate with Sentry. In the coming months you’ll start to see some of these changes in action. If you’re curious about the engineering behind the change, check it out in this PR. And if projects like this peak your interest, consider joining our team to get paid to work on them.