Your business · Practical guide
Build a dashboard for your business
Turn a small, known dataset into a useful dashboard with defined metrics, honest filters, and totals you can trace back to the source.
Before you start
- Pick one decision and no more than three metrics. Write their units, inclusion rules, and denominators so another person can calculate them.
- Create five fictional rows with known IDs and dates. Include a cancelled row and a different service; calculate expected totals outside the dashboard.
- Keep a separate fixture for duplicate IDs, blank amounts, invalid dates, and an empty file. Decide whether each problem rejects a row or the entire import.
- Start with a manual CSV snapshot. Decide who can see the data, where it lives, what source/freshness label appears, and how the previous version is recovered.
When this helps
Useful for a small operational view with a clear source and definitions you can check. Repeated spreadsheet reporting may only need the spreadsheet guide. Many sources, financial reporting, sensitive records, or important real-time decisions need a more established reporting system and review.
Build a small prototype first. Publishing, real user data, security and ongoing maintenance require separate checks.
Some setup and iteration · See the work involved
- Setup
- Start with a private browser project and one fictional CSV snapshot.
- Your work
- Define formulas, units, source ownership, refresh rules, and independently calculated totals.
- Iteration
- Reconcile filters and charts to source rows; test invalid imports, duplicates, reload, and recovery.
- Maintenance
- Own refresh, metric definitions, exports, access reviews, dependency changes, and hosting costs.
- One small, manually refreshed source with fictional records until checks pass.
- No paid connectors, forecasts, or financial-reporting claims.
An editorial estimate for this task, based on documented requirements. Your experience can differ.
Opening the brief builder…
Use your chosen tool
- Fill in the decision, source schema, exact metric definitions, filter rules, and refresh process. Prepare a brief and choose an eligible builder.
- Build an import preview and inspectable row table with the fictional dataset. Confirm validation and replacement behavior before creating charts.
- Implement one metric and compare it to your hand calculation. Add the remaining metrics only after the inclusion rules and rounding agree.
- Apply date and service filters, then reconcile each card and chart with the rows displayed. Check the start and end dates are inclusive as specified.
- Add an accessible weekly chart and responsive table, plus source label, last successful import, and explicit empty/error states.
- Reload and reimport the same fixture. Confirm totals remain stable. Try an invalid file and verify the previous valid snapshot is preserved.
- Verify private access and export/recovery with test data. Before real use, assign an owner for refresh, metric definitions, backups, access reviews, hosting costs, and changes.
An illustrative example
Written to show the intended shape of a result; this is not a measured tool test.
Starting material
Illustrative CSV fixture in one currency: J1, 2026-09-01, Design, completed, 120; J2, 2026-09-02, Design, completed, 80; J3, 2026-09-02, Design, cancelled, 50; J4, 2026-09-03, Support, completed, 60; J5, 2026-09-04, Support, pending, 40. Metrics count and sum completed jobs, plus average completed-job amount. Date filter covers all five rows.
What success could look like
Illustrative acceptance result: three completed jobs, total amount 260, average 86.67 rounded to two decimals. Filter to Design: two completed jobs, total 200, average 100.00. The cancelled and pending rows contribute nothing. An empty date range shows count 0, sum 0, and “No average.” Importing the same replacement snapshot twice must keep these results. This is manually calculated expected behavior, not a live builder test.
Check before using it
- Can every card be reproduced from the visible filtered source rows using the stated formula, units, and exclusions?
- Do cancelled rows remain excluded, and do blank amounts stay distinct from zero without changing the average denominator silently?
- Do the table, metric cards, and chart use exactly the same inclusive date range and category filter?
- Are duplicate IDs, invalid dates, and incompatible currencies or units surfaced before an import changes the valid snapshot?
- Does an empty filtered result show zero counts and “No average” rather than NaN, an invented trend, or stale previous totals?
- Does replacing or reimporting the same data preserve correct totals, source label, and last successful import information without duplication?
- Do reload and a failed import preserve the expected valid data, with a visible error instead of a false success?
- Can the intended owner see the records while signed-out or other test accounts are denied access to private data?
- Are chart labels, scale, table headings, keyboard controls, and small-screen views understandable without color alone?
If it needs work
- The total includes a cancelled job. Apply the same completed-only rule to the cards, chart, and table, then show the contributing record IDs.
- A second import doubled the totals. Implement the agreed snapshot replacement with a preview and reject duplicate IDs; retest the same file twice.
- The average uses every row even when filtered. Explain the denominator, use only the included completed jobs, and compare with the fixture calculation.
- This is a manual snapshot, not live data. Show the source and last successful import, preserve the previous snapshot on failure, and remove the “real time” claim.
- Remove the paid connector and automatic forecasting. Keep one fictional CSV source and three deterministic metrics until the checks pass.
The simpler alternative
Use a spreadsheet pivot table with a date and service filter, three checked formulas, and a small chart. A custom dashboard is worthwhile when repeated viewing or sharing justifies its refresh and maintenance work.
When a specialist product may be worthwhile
Use an established reporting tool and data specialist for many sources, large datasets, scheduled refresh, sensitive data, audited financial reporting, or business decisions that depend on dependable live data.
Sources and review
Documentation reviewed 2026-09-30. Plan limits and interfaces can change. How we prepare these guides.