---
title: "Build a dashboard from database questions"
description: "Build a Datarelix dashboard from database questions: review a result, create a chart, save it and refresh manually, with access and sharing limits explained."
summary: "Turn a checked sales result into a dashboard you can return to before the next reporting meeting."
canonical: https://datarelix.ai/resources/guides/build-dashboard-from-database-questions/
publishDate: 2026-09-10
updatedDate: 2026-09-22
lastVerified: 2026-09-10
---

# Build a dashboard from database questions

Turn a checked sales result into a dashboard you can return to before the next reporting meeting.

A useful dashboard starts with a question whose answer you can check. For a sales review, begin with **paid order value by month**, confirm the dates and amount definition, and then save the chart. In Datarelix, you do this through the conversation and refresh the saved work manually.

## What you need

Sign in to Datarelix and connect a supported source you are authorised to read. If you do not have connection details, ask the database owner for access to the reporting tables or views. The [quickstart](https://docs.datarelix.ai/quickstart/) covers both paths.

Each Datarelix dashboard uses one connection and is **private to its creator's account**. Team plans pool monthly questions across seats; they do not make dashboards shared. Confirm this fits the meeting workflow before building a report others need to open themselves.

## Ask for a defined result

For example, ask:

> Show paid order value in EUR for April, May and June 2026, grouped by calendar month using the order date. Exclude cancelled orders and label the amount before refunds.

Example result:

| Month | Paid order value, EUR |
|---|---:|
| April 2026 | 2,000 |
| May 2026 | 2,500 |
| June 2026 | 2,300 |

The total is EUR 6,800. Check that the query includes April 1 and excludes July 1, uses the intended time zone and sums each order once. If the result requires SQL knowledge to assess, ask an analyst to review it with the steps in [verify AI-generated SQL and results](/resources/guides/verify-ai-data-answers/) before saving the chart.

## Turn the table into a chart

Ask “Make this a line chart with month on the horizontal axis and paid order value in EUR on the vertical axis.” The plotted values should be 2,000, 2,500 and 2,300 in that order. Check the scale and units as well as the labels; a title containing “net revenue” would misdescribe this result.

Use the chart's [style controls](/features/#charts-title) if you need different labels or presentation settings. The underlying report still needs the same calculation checks. Styling cannot correct a duplicated join or missing filter.

## Add the chart to a dashboard

Open the chart's menu and choose **Add to dashboard**, then select a dashboard or create one. You can also ask “Add this to my Sales dashboard.” Add other checked charts from the same connection, such as orders by region.

Give each card a title that names its measure and period. You can reorder cards and edit dashboard card titles and labels. Those presentation changes do not rewrite the SQL; if the definition changes, revise the question and add the replacement chart.

## Refresh and interpret the result

Use the dashboard refresh control to rerun the stored query and rebuild its chart. **Refresh is manual.** Cards originating from the same question refresh together; **Refresh all** processes the dashboard's saved work.

If the saved query and chart cannot be reused, a refresh can rerun the original question with the model. Check the new result and query before relying on it.

The example above uses fixed dates, so refreshing it still reports April–June 2026. It may change if historical source records were corrected. For a rolling report, specify a relative period such as the previous complete calendar month, and verify the generated SQL uses that rule.

Refresh may incur database or warehouse charges. If it fails, check the connection and whether referenced columns still exist. Review the error before treating an older chart as current.

## Keep an analyst's review with the dashboard

For this example, the definition is paid order value before refunds and the period is April–June 2026. An analyst can compare the executed SQL with the [Snowflake version of this example](/resources/guides/natural-language-analytics-for-snowflake/) or the equivalent calculation for the connected engine. Confirm one order is counted once and that the timestamp follows the reporting convention.

When a business user exports the chart, include those definitions in the accompanying notes. When an analyst revises the report, distinguish a label edit from a query change. A chart still titled “April–June” should not silently become a rolling quarter, and a refreshed fixed-period chart may change after corrections to historical data.

## Bring the report to the meeting

You can download charts as PNG or SVG for a document or slide. Include the period, measure definition and the date of your check next to the export. The exported image is a **snapshot**; it will not update when the dashboard refreshes.

For the underlying report definitions, use the [weekly sales example](/resources/guides/weekly-sales-report/). The [dashboard documentation](https://docs.datarelix.ai/guides/dashboards/) explains the controls in more detail.