Insights

Overview

Insights is a powerful and flexible tool designed to visualize trends and compositions within your data. You can analyze events, cohorts, and user profiles, and display the data in a wide variety of chart types.

Advanced Insights features also allow you to create formulas, compare current data to past data, and generate custom events and properties for deeper analysis.

Use Cases

Here are some of the sample questions you can answer in Insights:

• Product Analytics
  • How is my WAU changing over time? (unique users)
  • How often are my users getting value? (frequency analysis)
  • What is the distribution of my users across regions / devices etc? (property breakdown)
  • Lifecycle analysis (opens in a new tab)
• B2B (in this case, a messaging application)
  • How many messages were sent in the US in the past 30 days? (total events, filtered)
  • How many users had a mobile app session yesterday? (unique events)
  • How many messages are sent per session? (formulas)
  • How much revenue was generated on plans purchased in the past year? (property aggregation)
  • How has the power users cohort grown over the past 6 months? (cohort trends)
• Marketing
  • Which advertising campaigns generate the most checkouts? (property breakdown)

Frequency Analysis

It's important to know what's the natural frequency at which your users use your product and experience the core value proposition of your product--do the majority of your users use your product daily? Weekly? Monthly? A16Z wrote a great article about the Power User Curve (opens in a new tab), and this video (opens in a new tab) shows how you can reproduce that within Mixpanel.

Quick Start

Building a report in Insights takes just a few clicks, and results arrive in seconds. Let's build a simple report together. Continuing the B2B messaging example, imagine you want to answer the following question:

Which cities in the United States have the most users who sent messages via the iOS platform?

Feel free to follow along and create your own report right in our demo project, here (opens in a new tab). To skip ahead and see the final result, click here (opens in a new tab).

Step 1: Choose Events

Events, cohorts, or profiles can be the basic building block of an Insights report. In this case, we want to know about users who sent messages, so within the "Events and Cohorts" section, add the "Send Message" event. At this point, your query should look like this:

Step 2: Choose Count Type

Next to your selected event, you can choose how to count that event. By default, Insights will count Total events, which, as the name implies, will count every occurrence of the event. In this case, we want to know how many users sent messages, so choose "Unique." Unique counts one event per user. At this point, your query should look like this:

Step 3: Choose Filters

Filters exclude unwanted data. In this case, we only care about events performed on the iOS platform. Therefore, add a "Platform" filter, where Platform equals "iOS Native". At this point, your query should look like this:

Step 4: Choose Breakdowns

Breakdowns segment data into groups. In this case, we want to count message sending users in different cities. Therefore, add a "City" breakdown. At this point, your query should look like this:

Step 5: Change Visualization

Choosing a different chart can help you visualize data better. Line charts help you see a trend, but other charts will help you see the aggregate value.

Step 6: Analyze Results

You now have a full analysis that you can use and add onto a Board, or use as a starting point to dig in further. This might mean adding more filters or breakdowns, or adding another metric and seeing if there's an interesting comparison to be made. You could also change the counting type or the chart type.

Basic Features

Chart Types

Insights features multiple visualizations to help you view the query results in the clearest chart type. By default, Insights displays the results on the line chart, which helps you understand how metrics trend over time. However, another chart type might present the results with more clarity. In Insights, you can choose to get a metric calculated either across the entire time period selected in the date picker or on a time-segmented view of the metric (e.g. daily breakdown).

• Metric calculated across the entire time period
  • Bar chart
  • Stacked bar chart
  • Pie chart
  • Metric chart
  • Table chart
• Metric time-segmented
  • Line chart
  • Stacked line chart

Sorting

Bar chart

When you view a bar chart, you have four different sorting options: A-Z Ascending, Z-A Descending, Value Ascending, or Value Descending. To switch sorting view, select the Events or property name column header in the upper left hand of the results and then select which sorting order you would like to see.

Line chart

Line charts in Insights are accompanied by a table of values that gives users another way to consume the trends information. This data table can also be sorted by clicking the column headers.

Click on a column header to sort by that column. Click the header again to reverse the sorting order. For example, the table below is sorted by event counts on August 2nd:

Results that are segmented (from one or more "group by" clauses in your query) have four different sorting options when you click on the "segment column" headers:

• Segment A-Z Ascending: sort by segment name in ascending order
• Segment Z-A Descending: sort by segment name in descending order
• Value Ascending: sort by segment value in ascending order
• Value Descending. sort by segment value in descending order

When sorting by segments, the sort is carried out left to right.

Clicking on the "Average" data column performs a flat sort across all segments:

Table chart

Tables are useful to see the precise values of your data and to quickly scan multiple metrics per segment. In general, tables work similarly to the rest of Insights, with a few extra features.

Sort Order

You can configure how you want rows in the table to be sorted, with our global sorting control.

Grouped View vs Ungrouped View

The Ungrouped View removes all hierarchy and makes the table flat. Each combination of segment values is treated as a row, independently of the other rows.

The Grouped View preserves hierarchy of breakdowns. It shows you segments within a breakdown as displayed below. This view is only applicable when you have 2 or more breakdowns

Alphabetical vs Value-Based Sorting

You can sort segments alphabetically or by the value of a particular metric. In the grouped view, sorting is configured on a per-breakdown level, and respects the breakdown hierarchy.

In the below image, we sort Country within Item Category, which respects the hierarchy.

Hierarchy is defined by the breakdown order in the query panel:

View “N” segments

The View “N” control lets you decide how many rows to display per breakdown in your table.

In the ungrouped view, choose the number of rows to display:

In the grouped view, you can choose the number of rows you want to display for each breakdown:

Notes

• View N only controls how many segments to display. To decide which segments to display i.e if it’s the top segments by value or bottom segments by value or alphabetical, please change the “Sort Order”.
• View N will show the minimum of N and how many segments are in your report. For example, if you set N=10 but you only have 7 segments, we will only show 7 rows.
• When selecting “Show All”, the maximum number of segments displayed in the UI will be 3000 given cardinality limits. If your breakdown has more than 3000 segments, you'll only see the top 3000, dictated by the sort order.

Overall and Segment Sub-Totals

Overall: This refers to the value considering all the segments, independent of whether displayed or not based on your View N control; i.e changes to View N will not effect Overall numbers

Sub-Totals: In the grouped view, in addition to the Overall, segment sub-totals are also displayed. Similar to “Overall”, these values are independent of the View N control

Analysis Settings

The Analysis options will determine the way the chart is calculated and visualized. The options are:

• Linear: This is the standard view for the chart.
• Logarithmic: A nonlinear scale based on orders of magnitude, rather than a standard linear scale, so the value represented by each equidistant mark on the scale is the value at the previous mark multiplied by a constant.
• Rolling: Rolling analysis calculates the rolling average of the data set. A rolling average curve is a series of averages from subsets of data. Use rolling average analysis to remove noise or spikes from data and smooth out trends over time. Mixpanel calculates the rolling average based on the selected time interval (hour, day, week, month or quarter) for each data point in the graph.
  For example, if you make a rolling analysis query for the past 30 days, Mixpanel calculates the rolling 7-day average by default. The value reported at each day in the line graph is the average of the values from the 7 days leading to that day. In the case of the first 6 days in your selected time period, the 7-day-average calculation will include days before the selected time period.
• Cumulative: Adds up the values of each point on the graph as it goes along, so the height of the line will increase over time.

Advanced

Formulas

Use Formulas to make calculations using simple arithmetic operators.

Mixpanel supports the following operators:

• + : Add
• - : Subtract
• * : Multiply
• / : Divide
• () : Use parentheses to influence the order of operations

Click the Formula button. Each event in the query shows a letter next to it, which indicates its variable name. Use these letters in combination with the operators to calculate a more advanced query. For example, you can use the DAU, WAU, and MAU functions in Formulas to calculate the stickiness of your product:

Enter a name for the formula (optional), and click Apply Formula to see the formula output. You can also use numbers as constants in a formula. Multiply a ratio by 100 to display as a percentage, for example. Divide a property value tracked in seconds by 3,600 to display the value in hours.

Custom Bucketing

Insights will automatically group your high-cardinality segments into buckets. Buckets can be edited by using the "Custom Buckets" option in the overflow menu: You can choose "Even" to get buckets of uniform size, and you can choose "Varied" to get buckets of different sizes. This helps with organizing outliers, or with drilling deeper into particular ranges.

Time Period Comparisons

Compare the current period of time to previous periods of time in order to track trends and growth in your product’s use. Compare traffic from a specific campaign period or event from one year to the next, or compare the success of that campaignto your normal traffic.

Note that if a data point for a previous year falls on a weekend, the data point is automatically moved to the next Monday to give a more clear picture of the data change from one year to the next.

Click on the Compare to past button at the top of your Insights graph and select the time period you wish to compare to. You can also select a custom date range.

Value Comparisons

When you have multiple metrics, or have broken down a metric by a property, you can compare them against each and the relative value. Click on Compare -> Overall. This works for all data functions. For a "totals" data function, you can see the percentage that a particular segment makes up. For non summable aggregations, it compares the segment value to the whole, unsegmented value.

Profile Analysis

Choose to explore either Events & Cohorts or Profiles. Events & Cohorts allows you to examine user behaviors, while Profiles allows you access profile data and visualize your users with filters and breakdowns based on their profile properties. When exploring Profiles, you are always analyzing all user profiles. Select the Data Function you want to use to calculate results by clicking on Total and selecting an option from the drop-down. You can calculate based on users or profile property value.

Annotations

In order to clarify the results in your Insights report add detailed annotations directly to the line chart. Annotations are tied to a specific date on the chart, rather than a specific data point on the chart.

Only project admins can create, save, and delete annotations.

To add an annotation, hover your mouse over the point on the chart you want to annotate, and click the blue + button that appears. Enter a description for the annotation, such as a holiday that occurred on that day or the end date of your fiscal year, then click Save. If you accidentally selected the incorrect date on the chart, you can edit the date and time of the annotation in this window.

View an existing annotation by clicking on the number found at the bottom of a report. You will be able to see who submitted the annotation.

Hover the cursor over the annotation to edit or delete it. Click on the pencil icon to edit an annotation, or the trash icon to delete an annotation. Add additional annotations to the same date by clicking Add annotation.

Insights Data Functions, Operators and Calculations

Data Functions in Mixpanel reports allow you to perform more complex calculations on your queries - this includes computing aggregate values of your event and property data, including totals, uniques, and averages.

The following data functions are only available in Insights, and are separated into groups based on what is being calculated: Total, Unique, Count users, aggregate property values, aggregate property values per user, and count sessions.

Select the Data Function you want to use to calculate results by clicking on Total and selecting an option from the drop-down. You can calculate based on events, users, event property value, event property value per user, and sessions.

The following functions have additional aggregation options which you will be prompted to select:

These functions provide additional aggregation options following the initial calculation because they are “per user” calculations. “Per user” calculations first calculate the value per user, which is an unhelpful query in its raw form, but becomes useful when you perform an aggregation on that calculation.

Selecting any of these functions gives you the option to choose different ways to aggregate this data. The default aggregation is Average, which you can click on to select a different option such as distribution, median, 25/75/90th percentiles, minimum and maximum.

If the data function you select calculates based on property (“Event Property Values” or “Event Property Values Per User”), you will be prompted to select an event property.

You can find more information about each data function and how they are calculated below:

Events

Users

Aggregate Property Values

Aggregate Property Values per User

Count Sessions

Additional Aggregation Option Examples

Here's a quick overview on "Distinct count" of property values, and how that differs from "Distinct count" of property values per-user:

Here are some use-cases that are now possible with "Distinct count" of property values:

• How many unique items were added to cart yesterday?
• How many unique songs / videos were played in the last 30 days?
• How many unique files were worked on in the last week?

Daily, Weekly, and Monthly Active Users

Daily Active Users (DAU): Select DAU to calculate the number of unique users in the previous day (24-hour) period that have performed the selected event. Weekly Active Users (WAU): Select WAU to calculate the number of unique users in the previous week (7-day) period that have performed the selected event. Monthly Active Users (MAU): Select MAU to calculate the number of unique users in the previous month (30-day) period that have performed the selected event.

If you select the DAU, WAU, or MAU function for a date range that includes the current day, the query will take the end of the current day as the end of the query’s time segment (even though it’s in the future). For example, today is April 25th, and it’s 4:22 PM. If you make a query to show WAU and you select “current day” as your date range, the query will return the count of unique users between April 19 at 12:00:00 AM and April 25 at 11:59:59.

View Users

Click a segment in an Insights report to see the list of users that underlie that data point. This helps see a representative sample of users from any analysis, so you can drill into anomalies or simply get to know your users. You can also save this user list as a cohort to either export or use for message targeting.

See this video walkthrough (opens in a new tab) for examples of how View Users can help you understand users contributing to a specific metric.

View Events

Click on a chart segment in Insights and view the raw events that made up that metric. You will be redirected to the Events page.

View Sample Events

You can hover over any event and in the context panel, you now have the ability to "View Sample Events", which redirects you to the Events page with 100 most recent samples of that hovered event: You can see a few samples of an event to help you decide whether that's the right event you want for your analysis or which property you should use for filters/breakdowns.