Inside Mixpanel

Introducing Automatic Segmentation

Emi Tabb

Most companies collect more data than they know what to do with and then struggle to extract useful insights that can improve their retention and conversion rates. As many Mixpanel users know, there are seemingly endless ways to segment a funnel – by country, device, platform, campaign source, signup date, and more.

For product teams, who report spending 72% of their time on tactical work, there simply isn’t enough time in the week to do that kind of deep data exploration. Fortunately, machines can crunch through billions of data points, very quickly, and uncover insights that might otherwise never be found.

That’s why we built our new machine learning model, Automatic Segmentation. It does the grunt work that data scientists don’t have the time to do, and searches through your most important funnel and retention reports to identify high and low performing segments of users. Then, it delivers that information directly to you, all in Mixpanel.  

This type of insight can directly impact how you build your user experience and where you spend marketing dollars. For example, GO JEK, a ride-sharing company in Asia, used the insights Mixpanel served them to alter their approach to product design.

“Automatic Segmentation spotted usability differences based on screen size for GO JEK users. This discovery inspired an entire research project which now impacts what we build for our users.”

        – Crystal Widjaja, VP of Business Intelligence at GO JEK.

Automatic Segmentation delivers needles from the haystack

When it comes to data exploration, many of us don’t know exactly what we’re looking for. That’s why Mixpanel will automatically segment each funnel and retention report that’s saved or on your dashboards by a variety of properties – both default and custom. Mixpanel then analyzes the findings and surfaces insights that deviate from the norm, helping users find that needle in the haystack. Does it provide answers to questions like: In which countries are users onboarding successfully at the highest rate? Does iOS or Android have better 2nd-week retention? Does usage vary by device or app version?

By doing the quick analysis that only a machine can do – deep but still wide in scope – Mixpanel can identify the optimal areas for further investment and improvement.

A curated list of Notifications across all machine learning models

When you have limited time, you want to spend it in the most impactful way. To help you out, we’ve built a global Notifications feed that rate limits and rank insights delivered by our anomaly detection and automatic segmentation models. As we add more models over time, this will become even more valuable.

If you’d like to see the full list of automatic insights, you can find them within your retention and funnels reports.

Our machine learning models also incorporate user feedback on whether each insight is relevant. So far, over 80% of automatic insights have been marked as “useful,” and we look forward to doing even better over time.

Gaining a competitive advantage in a world oversaturated with data.

According to IBM, in the last two years, we’ve created more data than the last 5K years of human history. But only 0.5% of that data is ever analyzed and used. To get through even a fraction of that data, product teams should supplement their people-power with machine-learning-equipped product analytics. At Mixpanel, we’re leading the charge.

**************

Mixpanel users already have access to Automatic Segmentation and can start seeing insights in their saved funnel and retention reports or in notifications.

Get the latest from Mixpanel
This field is required.