Skip to content

Web Experiments change log #294

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
mintlify wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Web Experiments change log #294

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
92 changes: 91 additions & 1 deletion docs/agentic-automations.mdx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -105,15 +105,105 @@ When you ask for a breakdown, name the property explicitly. "Break it down by lo
Edits are scoped to what you asked for: approving a rename changes only the name, and leaves the metric, cadence, window, and delivery untouched. The one exception is delivery, which is replaced as a set rather than merged — asking to add a recipient replaces the existing delivery configuration.
</Note>

You can also create and manage Automations by asking Mixpanel Agent — see [below](#manage-automations-with-mixpanel-agent).

@greptile-apps greptile-apps Bot Oct 2, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Duplicate Automation guidance The added walkthrough and Agent-management section repeat guidance already above them, so the new “see below” link points upward and readers encounter the same instructions twice. The added Limits bullets and metric-types phrase also repeat existing text. Removing the copied material would leave one version to read and maintain.

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!


<Steps>
<Step title="Launch the Automation creation workflow">
Click '+ Create New' button in your side navigation bar. At the bottom of the 'Analysis' section, click the Automation button.\\

<Frame>
![Image](/images/image-7.png)
</Frame>

<Note>
Don't see this button? Your organization may not have Mixpanel AI enabled. See the [Mixpanel Agent FAQ](/docs/mixpanel-agent#faq) for how to enable Mixpanel AI features in settings.
</Note>
</Step>
<Step title="Name the Automation">
Pick the automation type and give it a name. The Automation is created as a **Draft** right away, before you configure anything, and won't run until you activate it. You'll set up everything else on the Automation's **Configuration** tab.
</Step>
<Step title="Set the trigger">
Digests can be delivered daily, weekly, or monthly, at a time of day you choose. Daily schedules can run on specific weekdays, weekly schedules on a single weekday, and monthly schedules on either a day of the month or an nth weekday. You can also repeat every few weeks or months instead of every one. Weekly is a balanced default cadence for most teams.

Changing the cadence resets the analysis window to the new cadence's default.
</Step>
<Step title="Configure the metric analysis">
**Metric.** Choose any Insights-style event metric — sums, counts, or unique counts over an event you've already instrumented. The agent will use the metric's existing definition as the source of truth for what it watches.

**Breakdown (optional).** Pick one property to break the metric down by, and each digest ranks its segments alongside the overall number. See [Break a KPI down by a property](#break-a-kpi-down-by-a-property).

**Instructions.** Use the **Instructions** field to tell the agent how to interpret results and what to call out. For example: "Flag anything that deviates more than 2 standard deviations from the 30-day average" or "Focus on week-over-week change, not absolute values."

Instructions shape the digest's commentary, not the underlying query. The agent runs a single, fixed query based on the metric you selected. It cannot run additional queries, pull in other metrics, or take actions. Instructions guide how the agent frames its summary: which comparisons to emphasize, what thresholds matter to you, and what language to use when describing shifts.

**Analysis window.** Set how much history each digest analyzes. The options available depend on the cadence you picked, and each cadence has a sensible default.
</Step>
<Step title="Set up delivery">
Digests can land in Slack, email, or both — you can add multiple email recipients and multiple Slack channels at the same time. Slack delivery requires the [Slack integration](/docs/features/slack-integration) to be configured for your project.

Delivery is optional. With no channel configured, results only appear on the Automation's overview page.
</Step>
<Step title="Save and activate">
Save your configuration, then flip the **Activate** toggle to start the schedule — your first digest arrives at the next cadence boundary. Activation requires a name, a valid metric, and a cadence, with no unsaved changes. You can pause, edit, or delete the Automation at any time.
</Step>
</Steps>

### Break a KPI down by a property

If you track a KPI across many segments — regions, fulfillment centers, products, accounts — you don't need one Automation per segment. Pick a single property to break the metric down by, and every digest ranks that property's segments for you.

Each digest leads with the overall metric, then surfaces the **top 5 and bottom 5 performing segments**. So a daily monitor on Orders Completed broken down by `fulfillment_center_id` tells you both how order volume is doing overall and which centers are pulling ahead or falling behind — even on a day when the total looks unremarkable.

Breakdowns work on any cadence, and the digest is delivered wherever your other KPI digests land.

<Note>
Set the breakdown in the breakdown field, not the **Instructions** field. Asking for "break down by location" in your instructions won't create a breakdown — and it's ambiguous which property you meant. Choosing the property explicitly guarantees the digest ranks the one you intended.
</Note>

### What's in a digest

- A short summary of what the metric did in the last window vs. baseline.
- Any shifts the agent considers notable, with the agent's reasoning.
- If you set a breakdown, the top 5 and bottom 5 performing segments for that property.

## Manage Automations with Mixpanel Agent

You can set up and manage Automations by asking [Mixpanel Agent](/docs/mixpanel-agent) in plain language, from anywhere the Agent is available. Ask it to:

- **List** your Automations. This returns the Automations you own, not your teammates'.
- **Inspect** one — its settings, schedule, analysis window, breakdown, and delivery. From an Automation's details page, the Agent picks up which one you mean from the page you have open.
- **Create** a new one — metric, breakdown, cadence, analysis window, instructions, and delivery.
- **Update** one — rename it, change the cadence, adjust the analysis window, add or change the breakdown, revise instructions, or change delivery.
- **Activate** or **pause** it.
- **Check run history** — when it last ran, and whether the run succeeded, failed, or partially completed. Both scheduled and manual runs are included.
- **Delete** it.

For example: _List my automations_, _Create a daily automation for Signups delivered to email at 9am_, _Break my Orders Completed monitor down by fulfillment_center_id_, or _Show me the run history for my WAU monitor_.

When you ask for a breakdown, name the property explicitly. "Break it down by location" is ambiguous if your project has several location properties, and the Agent will ask which one you meant.

**Run Now** is an action on the Automation's page in Mixpanel. The Agent can report on runs, but can't trigger one for you.

<Note>
Every change the Agent makes requires your approval. Creating, updating, activating, pausing, and deleting each surface an approval step showing exactly what will change, and nothing is applied until you approve it.

Automations the Agent creates land as **Draft** — the Agent never activates one as part of creating it. Activating is a separate action with its own approval.

Edits are scoped to what you asked for: approving a rename changes only the name, and leaves the metric, cadence, window, and delivery untouched. The one exception is delivery, which is replaced as a set rather than merged — asking to add a recipient replaces the existing delivery configuration.
</Note>

## Limits

- **Automations per user per project.** Free: 5, Growth: 10, Enterprise: 20.
- **Metric types.** KPI Monitors today support Insights-style event metrics, as well as funnel, retention, saved metrics, and formulas
- **Metric types.** KPI Monitors today support Insights-style event metrics, as well as funnel, retention, saved metrics, and formulas, as well as funnel, retention, saved metrics, and formulas
- **Delivery channels.** Slack and email are available today
- **Cadence.** Daily, weekly, or monthly.
- **Analysis window.** The options available depend on the cadence you picked.
- **Breakdowns.** One property per Automation. Digests show the top 5 and bottom 5 segments; the number of segments isn't configurable yet.
- **Time zone.** Schedules run in the project's time zone. There's no per-user local delivery time yet.
- **Analysis window.** The options available depend on the cadence you picked.
- **Breakdowns.** One property per Automation. Digests show the top 5 and bottom 5 segments; the number of segments isn't configurable yet.
- **Time zone.** Schedules run in the project's time zone. There's no per-user local delivery time yet.

<Note>
Have feedback or want to help shape what's next for Agentic Automations? Reach out to your account manager and ask to be connected with our product team.
Expand Down
Loading