Practical framework / 5 MIN READ
Test what the dashboard does after a reader clicks
Verify filtering, highlighting and totals with a small expected-results dataset.
AI-assisted research and analysis for the SNS Creative Fieldbook
Sources checked 2026-11-06
A report can be correct in its default view and misleading after a filter or selection. Dashboard quality includes the interaction states readers actually use, the scope of each visual and the meaning of its totals.
What you will leave with: Build an interaction test matrix that catches hidden filters, inconsistent scope and non-additive totals.
Use this when
- Two dashboard cards disagree after a category is selected.
- A screenshot is circulated without its active filter context.
This lesson tests report behavior, not changing collection filters or granting access to restricted data.
Inventory the interaction paths
List date controls, category selectors, chart clicks, drillthrough pages and saved views. Note which elements should affect which visuals. A comparison card may intentionally remain fixed while a detail table changes, but that exception needs a visible explanation.
Microsoft documents distinct visual, page, report and drillthrough filters in Power BI, and distinguishes highlighting from filtering. Check the equivalent behavior in your reporting tool. A highlighted subset can leave the rest of the data visible, so a reader may misinterpret a total if the interaction is not clear.
Create a small fixture with known answers
Use a synthetic dataset with a few campaigns, dates and overlapping customers. Calculate expected spend, customers and rates by hand. Include a blank category, a zero result and a filter combination with no records. The fixture should be small enough that a reviewer can inspect every row.
Test totals according to the measure’s meaning. Spend may add across mutually exclusive categories; distinct customers may overlap; a conversion rate generally needs its own numerator and denominator rather than an average of displayed percentages. A grand total that differs from the sum of rows can be correct, but its logic must be explainable.
Exercise the full reader state
Open the report as an intended reader with approved permissions. Test the default, a single selection, multiple selections, reset, drillthrough and export. Check that hidden or persisted filters do not unexpectedly survive a reset or change the shared view.
Confirm that labels and empty states reflect the active scope. A no-data state should not appear as proven zero performance. Verify mobile or embedded behavior if those are real delivery surfaces. Do not bypass row-level access controls to make testing easier; permission behavior is part of the intended experience.
Release with a repeatable acceptance record
Record expected and observed values for the critical interaction paths. Treat mismatched scope or incorrect calculations as release blockers. Where a visual intentionally ignores a filter, label it and test that readers understand the distinction.
Attach filter context to exported results or ensure the export includes an adequate description. Keep the fixture and test matrix for later changes to data models or report controls. A refresh check proves the report loads; an interaction test proves that common reader actions preserve a coherent question.
Make the reasoning concrete.
A hypothetical dashboard has two campaigns with 10 customers each, including three customers present in both. The unique portfolio total is 17. A reader selects one campaign; its detail table changes to 10, but an unlabeled fixed portfolio card remains 17.
Force every total to add
Opening: Change the portfolio card to 20.
Hypothesis: The displayed rows sum neatly.
Risk: Three people are double counted.
Clarify and test scope
Opening: Keep 17 as unique portfolio customers and label any fixed card; filter other views as specified.
Hypothesis: The reader understands both overlap and interaction.
Risk: The design still needs a usability check.
Use a fixture to verify the distinct-count logic and the selected-state behavior. Do not repair a scope confusion by corrupting the measure.
Take the questions into your next brief.
Use these prompts in your own notes. Nothing is submitted from this page.
Interactions
Which controls should affect each visual?
Fixture
What small dataset has known answers?
Totals
Which measures are additive and which require recomputation?
States
What happens after reset, drillthrough and export?
Evidence
Which expected-result checks block release?
Watch for the wrong conclusion.
- Testing only the default view.
- Averaging displayed rates to produce a total.
- Summing overlapping unique customers.
- Sharing a chart without active filter context.
Your next action
Create a ten-row synthetic dataset and test the five most common reader interactions against hand-calculated results.
Sources & scope
Frameworks, worksheets and hypothetical examples are original editorial synthesis. Linked primary sources support only the stated concepts. Source checking is not an independent audit of effectiveness. No named human author or reviewer is claimed.
- Microsoft Learn: Filters and highlighting in Power BI reports →Primary technical documentation · Supports: Documents visual, page, report and drillthrough filters and distinguishes filtering from highlighting. Does not certify any particular report’s calculations or permissions.