Introducing Anomaly Detection and Root Cause Analysis

We love when customers make it a habit to check key metrics in Mixpanel every morning. Who doesn’t get a little excited to thumb through reports and check in on the progress of the latest marketing campaigns or product feature launches?

But let’s face it, staying on top of all your metrics all the time is really hard.

That’s why we’re excited to offer a set of features designed to save you time while keeping a close eye on performance. With Custom Alerts and our new Anomaly Detection and Root Cause Analysis features, you can stay on guard and get ahead of potential issues (or get flagged to a sudden metrics win), without all the manual effort.

Catching and diagnosing regressions

You never want to be caught on your back foot by a leader noticing a dip in metrics before you do. That’s why we developed simple alerts that monitor fluctuations in performance and quickly send out notifications in your important work streams.

In one fluid workflow, you can set to have your team alerted via Slack or email when a metric hits either a custom threshold or an automatically calculated threshold.

Anomaly Detection, the latter, auto-calculates a threshold according to a metric’s expected range, without needing you to set any specific values. We leverage the open-source tool Prophet to forecast your time series data and determine whether the data point falls outside of your selected confidence interval. If you’re interested in dialing in a custom threshold you want instead, you can do it with Custom Alerts.

How Anomaly Detection works:

Once you’ve received an alert, your next question is typically: “What happened?”

With any fired alert, you can create a Root Cause Analysis in one click that will kickstart your investigation of what might have caused the metric change. Our Root Cause Analysis is an automatically generated Board with 12 reports broken down by popular properties of the event.

How Root Cause Analysis works:

Here are some potential scenarios to show how these alerts and Root Cause Analysis work in Mixpanel:

• You’re alerted in the #marketing Slack channel of a major drop in traffic for a key landing page. From the Root Cause Analysis, see from a last-touch breakdown that there’s an issue in a paid ad campaign no longer driving traffic.
• An email hits your inbox to notify you of a major drop in your team’s recent feature launch engagement. From the Root Cause Analysis, you see in the browser breakdown that there’s a severe drop-off for Safari due to a new bug.
• The #engineering Slack channel gets a ping about a major drop in DAU. From the Root Cause Analysis, see that the drop is driven by users in APAC, coinciding with the Lunar New Year holiday.

Catching issues is also often about teamwork. We’ve designed our system so that each alert is linked to only one Root Cause Analysis so you aren’t duplicating work if others have already diagnosed the problem. Click into a Root Cause Analysis, see that a teammate is already on it, and rest easy.

Celebrating wins

We’ve talked a lot about potential issues, but it’s not all about negativity. Alerting is also useful for celebrating positive moments.

It’s easy to lose sight of the bigger picture when your team is deep in building mode every day. Alerts provide a great opportunity to celebrate progress and keep the team motivated by seeing that their work is moving the needle.

Here’s an internal example of how we use alerts for good:

Get started

Anomaly Detection and Root Cause Analysis are available on Enterprise plans today. For more information, check out our docs.

Own your metrics, diagnose changes, and move fast. Staying on top of your work is easy with Mixpanel 🤝

Gain insights into how best to convert, engage, and retain your users with Mixpanel’s powerful, self-serve analytics. Try it for free.