Turn your data into a dashboard that helps you understand it
Define the question and metrics, check calculations, then create clear charts and filters.
Let's understand the idea
Start with the question the dashboard should answer: what changed, where, and over which period? Then choose and define metrics, instead of filling cards with numbers from unclear sources.
Prepare a clean copy of the data and request calculation review before designing charts. Clarify whether it's a static file or an actually connected source. A refresh button without a source isn't a live update. Test filters and the no-data state, and check totals against an original sample.
Tools for this workflow
Choose from what's available to you. You can use these prompts with several tools. What they can carry out depends on supported features and files.
See tool details and uses in the Tools library →Examples to expand the idea
A clinic manager has a spreadsheet of the whole year's appointments. She uploads it to Claude and requests a dashboard: monthly appointments, busiest days, and the percentage who booked but didn't attend. She checks one month's figure manually before presenting it.
YouTube or content performance
Start with an export whose column definitions you understand. A direct connection needs permission and a service supported by the environment.
An expenses dashboard
Distinguish dates, currencies, categories, and amounts. Review numbers extracted from invoice images before relying on them.
An operational metrics dashboard
Choose metrics that call for decisions. Show missing data instead of filling it with invented values.
Steps from start to delivery
Open any step by its heading. Each explains what to do, why, and what to check before continuing.
1Write the decision questionA few metrics with a purpose.Details
A dashboard helps understanding, beyond its appearance.
What should I do?
- Define the reader, question, and time period.
- Define metrics and their calculation methods.
2Inspect the data tableData you can verify.Details
A missing value or different unit can change the meaning.
What should I do?
- Check columns, dates, units, and duplicates.
- Keep the original and record cleaning rules.
3Approve calculations and chartsApproved calculations and a justified chart choice.Details
A chart must represent numbers without misleading implications.
What should I do?
- Reconcile totals, percentage changes, and groupings.
- Choose a chart for the question, with clear units and titles.
4Build the interactionsA dashboard that responds to the data.Details
Filters should actually change the numbers.
What should I do?
- Try changing the time period, category, and table.
- Test no-data states, source errors, and chart values during interaction.
Tools for this part: choose oneA guide for this stepHow do you build a website or app with AI?5Define updates and sharingA clear update method and appropriate access.Details
Old information can be mistaken for current information.
What should I do?
- Show the source date and how it's updated.
- Try a new file. For private data, review the sharing audience and permissions.
Prompts for your work
Open a prompt and fill in your details. Empty fields remain visible so you don't forget them.
Prepare the dashboardOpen it to copy or tailor the prompt
Prepare the listed inputs and start with the prompt in an available tool. Open the output and review it against your work.
Copy the prompt as it is, or tailor it with your details first.
Tailor the prompt to your workOptional: add your details and the prompt adjusts itself
View full prompt
Why is the prompt written this way?
Clarifying behavior and inputs helps us build something people can use, beyond just a page's appearance.
Test the dashboardOpen it to copy or tailor the prompt
Copy the prompt as it is, or tailor it with your details first.
Tailor the prompt to your workOptional: add your details and the prompt adjusts itself
View full prompt
Did the result serve your original goal?
Return to the inputs and sources, and open the output where you'll use it. Check these points before relying on it.
- Do calculations match the source?
- Do filters change the result correctly?
- Are the update date and privacy clear?
If there's a problem, name the unsuitable part, show an example, and request a specific edit.