From Question to Dashboard

Dashboards should be the output of trusted analysis, not the place where ambiguity goes to hide.

The right sequence

The healthiest dashboard workflow in Tukun.ai is:

1. ask the question in the Workbench
2. refine the answer until the result is conceptually correct
3. model the metric if the meaning is unstable
4. save the stable output as a card
5. group related cards into a dashboard

If you reverse that sequence, you often end up with polished dashboards built on unresolved metric debates.

What belongs on a dashboard

Put a result on a dashboard when:

• the question recurs on a regular cadence
• the metric definition is stable enough for reuse
• multiple people need the same view
• the card has already survived review

What does not belong on a dashboard yet

Keep a result in the Workbench when:

• the business meaning is still contested
• the query shape is still changing often
• the source data is unstable
• the chart answers a one-off investigative question

A good dashboard design principle

One dashboard should support one operating conversation.

Examples:

• weekly product usage review
• monthly revenue review
• acquisition channel review
• retention and expansion review

That keeps the dashboard scoped, maintainable, and easier to trust.

A practical promotion checklist

Before promoting a card into a dashboard, confirm:

• the title is understandable outside the original conversation
• the metric definition is stable
• the filters are intentional
• the source is expected
• the card would still make sense a month from now

The real goal

The goal is not to create more dashboards. The goal is to make the right decisions easier to revisit with less argument about what the numbers mean.