Metric Tree

Overview

Mixpanel Metric Tree is a “live” logical decision-making framework used to represent relationships between metrics and show the flow between inputs and outputs. The “live” metrics make it possible to understand the quantitative impact and trace the inputs driving it.

It is one space that simultaneously depicts both quantitative and qualitative data along with additional context, making it possible to discuss insights as a team across functions and make data-driven decisions together.

Use Cases

• Data-driven decision: Clearly articulate what’s working vs not by looking at how your input metrics and initiatives are impacting your top business outcomes
• Align with your team, from Execs to ICs, and across teams on growth strategy & metrics by: Use this as the codified strategy map
• Collaborate on strategy, insights & actions during team rituals: Leverage this during weekly rituals to identify which areas require more focus, and add those insights in the tree for others to reference
• Evolve your metric tree, as your strategy evolves: With every planning cycle, update the initiatives and input metrics as goals and focus outcomes change.
• Use this as a plan to ensure you have the right data tracked: Use this as a guide to let you know which additional data needs to be tracked, or to ensure it’s modeled in the right way to define the metric

Quick Start

Step 1: Create your Metric Tree outline without data

The idea is to think through your strategy and the various hypotheses without confusing it with the data available today or the current value of metrics. Your strategy is more long-term. Use this strategy map to discuss and align with your teammates before putting in the effort to connect to a metric with data. To create a metric tree outline

• Add empty state metric cards to the tree. To create a metric card, choose Metric from the toolbar below.
• You can name these metric cards and link the connections. You can also add strategy cards to show the initiatives in place

Step 2: Make the metrics real with data

Click on connect metric, and connect the metric card to the metric with data. This is a regular Mixpanel query panel. You can choose to bring in saved metrics, or create a metric.

Step 3: Add more context to your metrics

Make your tree more actionable for yourself and the team by adding context to it - whether it is pinned content with some interesting deep dive reports, or examining correlation and finding interesting insights to add to the logbook. You can learn more here.

Step 4: Share your metric tree & get started!

Metric Trees have similar permissions to Mixpanel boards. You can share the tree with anyone in the project: with view or edit access.

• Editors get access to edit the definition of metrics, add new metrics, add context, etc.
• Viewers get access to open and see the underlying formula definitions, and read the context and notes

Metric Tree is a multi-user space meant for collaboration and for you to use during your weekly team rituals. So many of you can be on the tree at the same time and edit it, and most importantly, add insights or actions for metrics you’re accountable for.

Customize each metric

Time comparisons that are relevant to the metric

To ensure your metric is showing the most apt tome comparisons, you can now customize at a metric level which time comparisons do you want to show in the metric tree view.

For example, if I wanted a QoQ view instead of YoY, I would select:

• Time range: Last 3M
• Compare to: Previous quarter
• Interval: week; Interval determines the number of data points you have displayed in the metric line chart

Add owner to keep a person accountable for the metric

For any eligible metric, you can click on the owner icon and assign an owner

Pin contextual reports & boards to the metric, to enable deep-dive

Click on a metric. When the query panel opens, navigate to the Context tab. Within the Context tab, pin reports and board in the ‘Pinned’ tab

Add important insights to the logbook for future reference

Click on a metric. When the query panel opens, navigate to the Context tab. Within the Context tab, add entries into the ‘Logbook’ tab

FAQ

What does the “As Of” control on the top left of the metric tree do?

As of determines the end date reference point for the metric tree.

For example, if you want to look at how the metric tree looked at the end of Q1, i.e, March 2025, you could set “As of 31 March 2025”. All the metrics in the metric tree which are based on last 30 Days, or the last 3 months etc, will look at measuring the value of last 30D as of 31 March 2025, i.e, 2-31 March 2025, instead of last 30D as of today.

What does Filtering a tree do?

Filtering a tree is similar to filtering a board. It filters every metric in the metric tree. If you’re looking to filter only a specific metric, please use inline filters.

How do I see the metric trend chart (metric time series)?

Access the 3-dot overflow menu on the top right of the metric tree. Click on Show/ Hide > Show Metric Chart.

If you want to look at a longer time series chart for a particular metric, go to the specific metric card, click on the overflow menu, and choose View Chart.

I want to move trees across projects. How do I do that?

You can copy a tree and paste a tree in a new project. To select the whole tree, use the select tool in the toolbar below, in the middle of the screen. Then Copy (Cmd + C), and Paste (Cmd + V) in the new project.

How many trees do you suggest we create?

Create minimal trees so there is a complete map of the strategy in one place vs across disparate trees. Ideally, start with one tree per company, and depending on the size, maybe create separate trees for each function or business line if required

Does the top of a metric tree have to be revenue-related?

No, it doesn’t have to. We generally ask people to start a tree at L2 or L3 level where there is more functional ownership, for example, marketing can own Volume of Sign-Ups or Product can own WAU (Weekly Active Users).

What is correlation? How do I access it?

Access the 3-dot overflow menu on the top right of the metric tree. Click on Show/ Hide > Show Correlation. Correlation numbers should show up across all the edges, i.e the connection between two metric cards.

Mixpanel shows Pearson Correlation Coefficient. It measures the strength and direction of a linear relationship between two time-series data sets. It produces a value between -1 and 1

For two metrics, A and B, each return time-series data, we compute Pearson correlation using the following formula:

$r = \frac{\sum{(A_i - \overline{A})(B_i - \overline{B})}}{\sqrt{\sum{(A_i - \overline{A})^2} \sum{(B_i - \overline{B})^2}}}$

Where:

• $A_i$ and $B_i$ are individual values in the time-series data of metrics A and B.
• $\overline{A}$ and $\overline{B}$ are the mean (average) values of metrics A and B over the observed time period.
• $r$ is the Pearson correlation coefficient.

The resulting r-value determines the correlation score displayed on the edge.

Interpreting the Correlation Score

• (0.6 to 1.0) Strong Positive Correlation: The two metrics tend to move in the same direction strongly. As one metric increases, the other increases. Similarly, as one metric decreases, the other decreases
• (0.3 to 0.6) Moderate Positive Correlation : Some relationship exists
• (0.0 to 0.3) Weak Positive Correlation : Some relationship exists, but it is not strong
• (0.0) No Correlation: Changes in one metric do not seem to have changes on the other
• (-0.3 to 0.0) Weak Negative Correlation : Some inverse relationship exists, but it is not strong
• (-0.6 to -0.3) Moderate Negative Correlation : Some inverse relationship exists.
• (-1.0 to -0.6) Strong Negative Correlation : The two metrics strongly move in opposite directions. As one metric increases, the other decreases and vice versa.

To compute the Pearson correlation, we use:

• Time-series data from both metrics.
• Matching timestamps: Data points are aligned by timestamp to ensure proper comparison.
• A configurable time window: The correlation is typically calculated over a specified time range (e.g., the last 30 days)
• Some specific aspects to call out:
  1. If 2 time-series data have different lengths, correlations are calculated based on the overlapping lengths between the 2
  2. All correlations are run on a daily granularity, since event data is available on a daily basis to compute metrics at that granular view.

Limitations of Correlation

• Correlation does not imply causation: A high correlation does not necessarily mean that one metric influences the other.
• Non-linear relationships: Pearson correlation captures only linear relationships. If the metrics are related in a non-linear way, this score might not reflect the true relationship.
• Data Quality Matters: Missing or misaligned data points can affect the accuracy of the correlation score.