Dashboards

Dashboard overview

Dashboards are persistent analytical pages for recurring reporting. They can be created by users, edited with the Agent, or built from saved chart artifacts.

[picture 1. Dashboards page showing published dashboards and private drafts]

Dashboard states

Stackless dashboards can be:

• Private drafts that belong to a user or conversation.
• Published dashboards shared with the organization or specific roles.
• Forks created from a published dashboard for safe editing.
• Versioned dashboards with snapshots that can be reviewed or restored.

Dashboard building blocks

A dashboard can include:

• KPI cards.
• Line, bar, and other chart widgets.
• Table widgets.
• Tabs or sections.
• Date filters.
• Dimension filters such as location, channel, product, or customer segment.
• Metric definition drawers.
• Scheduled exports.

[picture 2. Example dashboard with KPI cards, trend chart, filter bar, and table]

Creating a dashboard with the Agent

The Agent can create a dashboard from a request such as:

• "Build a dashboard for weekly revenue, orders, AOV, and active customers."
• "Create an ad performance dashboard by channel and campaign."
• "Add a tab for customer retention and a tab for acquisition."
• "Use these chart cards and make a single executive dashboard."

The Agent will look for existing Semantic Models and Transformation Models first. If a model is missing, it may propose the needed model work before it can build the dashboard correctly.

Editing and publishing

For private dashboards, edits can be made directly. For published dashboards, Stackless uses a safer edit flow:

1. Fork the published dashboard into an editable draft.
2. Make changes in the draft.
3. Preview the dashboard and dependent data.
4. Publish the replacement when it is ready.

Publishing can also promote owned draft dependencies, such as draft transformation or semantic model work, when that is part of the reviewed publish plan.

Filters

Dashboard filters should be explicit. Common filters include date range, location, channel, campaign, product, customer segment, and region.

If a dashboard mixes widgets from multiple Semantic Models, filters should be bound to compatible members. Otherwise a filter can appear to work for one widget while not applying to another.

Metric definitions and lineage

Use the metric definition drawer or ask the Agent:

• "Explain how this KPI is calculated."
• "Trace this metric back to source tables."
• "What grain does this widget use?"
• "Which transformations feed this dashboard?"

This is the fastest way to resolve metric-definition disputes.

Scheduled exports

Dashboards can be exported or scheduled when the workspace has export permissions enabled. Scheduled exports are useful for recurring executive packets, weekly operations reports, or account-team updates.

Use Monitor and the scheduled export drawer to inspect run status, delivery history, and failures.

Good dashboard prompts

• "Create an executive dashboard with revenue, orders, AOV, active customers, and a weekly trend."
• "Add a table by channel with spend, revenue, ROAS, orders, conversion rate, and new customers."
• "Make location a filter across all widgets."
• "Split this into Overview, Customers, and Marketing tabs."
• "Explain every metric on this dashboard and show the upstream models."