← Back to Knowledge Base
Affect Performance Team
|CRO|Nov 22, 2026

Mixpanel Updates: Custom Insights Query API, Sample Size Calculator, Experiments on Boards and Claude MCP

From October 15 to November 3, 2026, Mixpanel introduced four updates to its analytics and experimentation tools. The platform added the Custom Insights Query API for retrieving analytics results programmatically without first creating reports, a built-in sample size calculator for A/B testing, experiment results on shared Boards, and a centrally managed connection to Mixpanel for employees using Claude.ai.

The updates cover several stages of working with data, including building analytics queries, planning experiments, reviewing results with colleagues, and using AI assistants. Mixpanel is making its analytics available through APIs, automated workflows, and centrally managed AI access, expanding on the reports and manual workflows in its interface.

Custom Insights Query API for Analysis Without Saved Reports

On October 28, Mixpanel introduced the Custom Insights Query API in beta. The new endpoint returns analytics results directly to applications or scripts.

Previously, to retrieve Insights results through the Query API, teams first had to create a report in Mixpanel, save it, and query it by ID. Changing the metrics, filters, or time range meant editing the saved report or creating additional versions. The new endpoint accepts query parameters in the request body and returns the results without requiring a report in the Mixpanel interface.

Supported Metrics, Filters, and Output Formats

The Custom Insights Query API supports several types of metrics, including events, funnels, retention, cohorts, and formulas. Users can set a date range and time granularity and add up to 10 filters and three breakdowns.

Breakdowns can use event properties and cohort membership. The API also supports comparisons with an earlier period and can show how results changed relative to a chosen baseline. It returns results in JSON or CSV for use in custom applications, automated reports, and analytics workflows.

For example, a SaaS company can retrieve its signup count, user activation rate, and paid subscription conversion rate for the past seven days on a daily basis. An application sends a query with the required metrics and date range, then uses the results from Mixpanel to build a report or detect anomalies. Marketing teams can also query campaign results by UTM parameters, cohorts, and time ranges without saving a separate report for each channel or segment.

Setting Up the Custom Insights Query API

The endpoint uses a Mixpanel service account for programmatic access. Developers can create a service account or use one with the necessary permissions, then submit a request with the parameters described in the Custom Insights Query documentation. The Mixpanel-Version header sets the API version, and the application processes the results in the selected format.

This lets teams build analytics queries into automated workflows without maintaining multiple saved reports. Mixpanel introduced the API in beta, so teams should check the authentication rules, supported parameters, and limits for the version they use.

Sample Size Calculator in Experiment Setup

On October 15, Mixpanel added a built-in sample size calculator to experiment setup. It estimates how many users a test needs and how long it should run to detect a change in the selected metric.

Teams could previously use a separate Mixpanel calculator, but they had to look up and enter the baseline conversion rate and traffic manually. They can now run the calculation during experiment setup, and Mixpanel can fill in historical data from the relevant project.

How the Calculator Works

The user selects a primary metric, such as signup conversion, completed purchases, or feature activation. Mixpanel uses the metric’s historical values and the number of eligible users to estimate the sample size and test duration. The calculator uses the following inputs.

  • Baseline. The starting value of the metric, such as the current conversion rate.
  • Minimum Detectable Lift. The smallest change the team wants to detect.
  • Confidence Level. The chosen level of statistical confidence.
  • Statistical Power. The probability of detecting an effect of the specified size if it exists.
  • Rollout Percentage. The share of users who participate in the experiment.
  • Hypothesis. A one-sided or two-sided statistical hypothesis.

The calculation also accounts for the number of variants and the primary metrics selected. For example, an online retailer testing a new layout at the top of a product page can use add-to-cart conversion as its primary metric. Mixpanel uses historical conversion and eligible traffic to calculate how many users the retailer needs to detect the specified minimum lift.

This helps the team decide whether it can run the experiment within the desired time. If the site has too little traffic to detect a small change, the team can adjust the test parameters or choose a different approach.

Using the Calculator

The calculator is available in Experiments settings. After choosing the primary metric, the user selects Calculate sample size next to the duration settings. Mixpanel fills in available historical values, and the user can adjust the minimum detectable lift, statistical power, confidence level, and other inputs.

The calculator shows how many users the test needs in total and per variant, along with the expected duration. The team can apply the estimate to the experiment settings with Apply to Experiment.

The estimate helps teams plan around actual traffic and the expected effect. It cannot guarantee statistical significance or tell the team which variant will win before the test runs.

Experiments on Boards Alongside Product Metrics

On October 21, Mixpanel added the option to show experiment results on Boards, the shared dashboards that bring together reports and other items. Experiment results previously lived mainly on the experiment page. To show them next to product metrics, teams could share a separate link or copy metrics and screenshots into a shared dashboard.

Teams can now add an experiment to a Board as a card. The card shows the main test results in the context of other business metrics.

What Experiment Cards Show

Users can choose what appears on the card. Options include primary metrics and lift relative to the control, exposures by variant over time, descriptions and images of the variants, and the original hypothesis.

For example, a SaaS product team can use one Board to track signups, activation, and retention alongside results from an experiment with a new onboarding flow. The team can see how product metrics are changing, how large the experiment is, and how the test results relate to current product trends.

Teams can add an experiment card to several Boards. The same test can appear on both a product team’s Board and a shared leadership report.

Adding an Experiment to a Board

Teams can add an experiment from either the experiment page or the Board. The options are listed below.

  1. Open the experiment page and choose Add to board from its header menu.
  2. Open the relevant Board and add an Experiment card.

After adding the card, the user chooses which experiment metrics and details to show. The cards are view-only. To edit the experiment settings, select View experiment to open the original test.

That makes experiment results part of regular analytics reporting. Teams no longer need to maintain separate screenshots or move metrics between reports by hand.

Mixpanel MCP Connects Employees to Claude.ai Through a Central Setup

On November 3, Mixpanel added Managed Authorization for Claude.ai to its MCP Server. The update lets administrators connect Mixpanel to Claude centrally for employees in their organization.

Previously, each user had to connect the AI assistant to Mixpanel, sign in, and approve permissions. That made the integration harder to roll out and manage at larger companies. Administrators can now configure the connection once, and employees can use Mixpanel in Claude without a separate sign-in or consent screen on first use.

Working With Mixpanel Data From Claude

The MCP Server connects the AI assistant to Mixpanel’s analytics tools. Users can ask about product metrics in natural language and get answers based on the data their account can access.

For example, a product manager can ask Claude how signup conversion changed over the past week, compare retention across user cohorts, or request a report on user behavior after a feature release. Users describe the task in the conversation instead of setting up each analytics report manually. Claude calls Mixpanel’s tools and uses the results to answer.

Users can also work with reports, cohorts, and other Mixpanel objects through MCP within their assigned permissions. Claude gives users another interface for working with the data already in Mixpanel.

How Managed Authorization Works

Managed Authorization requires setup by both a Claude administrator and a Mixpanel organization administrator. In Claude, the administrator uses the connector’s Managed authorization section. In Mixpanel, the administrator goes to Organization Settings → Access Security → Claude Access.

The administrator can choose between two identity providers. Claude can act as the identity provider and simplify initial setup. An Okta integration lets the organization use its existing corporate sign-in and multifactor authentication policies.

After setup, staff retain their existing Mixpanel permissions. Claude acts on behalf of the individual user, and Mixpanel checks their access to projects and data on each request. Managed Authorization grants employees no additional permissions.

That matters for organizations where teams can access different projects, customer data, or areas of analytics. For example, marketing staff can use the campaign metrics available to them, while product teams can use behavior data for their own products.

Connection Auditing and Access Requirements

Mixpanel records new connections in its audit log. A successful authorization generates the mcp_client.authorized event, and changes to trusted identity provider settings appear under trusted_identity_issuer.* events. Administrators can use the log to track AI tool connections and changes to corporate access settings.

Managed Authorization is available in beta on all Mixpanel plans, including Free. It requires Claude Team or Enterprise, an active MCP Server for the Mixpanel organization, and a verified corporate email domain that matches the domain of employees’ Claude accounts.

Central setup simplifies connecting users while preserving Mixpanel’s permission checks. Teams still need to manage who can access the underlying data.

Mixpanel for Product Intelligence and User Behavior Analysis

Mixpanel combines product, web, and mobile analytics, session analysis, metrics management, experimentation, and AI tools in one Product Intelligence platform. Product, marketing, and analytics teams use it to understand user behavior, investigate why key metrics changed, and test changes intended to grow the business.

Mixpanel’s analytics are event-based. In addition to page views, it analyzes actions within digital products, including signups, feature views, add-to-cart actions, transactions, and return visits. Teams can build funnels, measure retention, compare segments, and examine the sequences that lead to purchase, activation, or abandonment.

Mixpanel combines quantitative reports with Session Replay and Heatmaps to help teams investigate user behavior. A team can spot a drop in funnel conversion and then review session recordings to see where users struggled. This connects aggregate patterns with observations of individual interactions.

Experiments and Feature Flags let teams roll features out gradually, create test variants, and measure the effect of changes on product metrics. Experiments use existing events, metrics, and cohorts, giving teams a shared set of data for planning and evaluating tests.

Metric Trees connect product metrics to business goals. For example, a company can use activated accounts, retention, usage of key features, and revenue as related levels in a measurement model. This helps teams see how a change in one metric affects broader business results.

Mixpanel is also developing AI tools for natural-language analytics. Mixpanel Agent helps users explore data, build reports, and diagnose metric changes. Mixpanel MCP makes the platform’s tools available to external AI clients, including Claude, ChatGPT, and other compatible tools. Connections to cloud data warehouses and other systems let teams use data from their existing infrastructure.

Marketing teams can use Mixpanel to see what people do after clicking an ad, including whether they sign up, activate, use the product, buy, or come back. Product teams can use the same analysis alongside feature management and experiments. Leadership can connect operating metrics to growth goals.

The October and November updates extend these capabilities. The Custom Insights Query API makes analytics results available to software, the calculator helps plan experiments, Boards put test results alongside product metrics, and Managed Authorization provides centrally managed access through Claude. Together, these changes make it easier to measure behavior, investigate its causes, and test changes to a digital product.

Sources

  1. Mixpanel Changelog, announcements from October 15, 21, and 28 and November 3, 2026
  2. Mixpanel Custom Insights Query API, programmatic analytics queries
  3. Mixpanel Experiments, sample size calculator and statistical settings
  4. Mixpanel Claude Enterprise Managed Authorization, connecting Claude across an organization
  5. Mixpanel Product Intelligence Platform for the AI Era
  6. Mixpanel Experiments and Feature Flags
  7. Mixpanel Session Replay and Heatmaps

Monthly Overview

Stay ahead of the market. Every month.

Our practitioners pick the TOP 10–20 news from the world of Paid Media and MarTech for the month.

10–20Top news / month