Review calculations, accessibility and refresh claims
Learn the reasoning. Try the task. Compare your answer.
Prepared for DataCareerHub · Curriculum version 2026-10-15.5
By the end of this lesson
- Test source-to-dashboard reconciliation and interaction states.
- Check readable alternatives and keyboard use.
- Document fixture freshness and a practical handover process.
Verified April–May source from Module 4 · EUR · Static offline practice · Open the worked dashboard ↗
Understand the method
A dashboard is ready for review when its calculations, interactions and explanation have been checked together. A screenshot verifies only one visual state. Use a small test matrix to trace source values through cards, charts and tables, then inspect how the narrative changes. The supplied expected-state file is a learning comparison, not a substitute for defining your own analytical contract.
Check the calculation chain
Reconcile completed-order counts and values against the verified source. Across both months, there are 40 completed orders totalling €6,320, with AOV €158. Returned and cancelled orders stay outside those measures. Count the underlying table rows for selected views and confirm that channel values sum to the same eligible total. Recalculate ratios from their components rather than averaging monthly percentages or unweighted group means.
Test the difference between missing and zero. A blank delivery duration is unavailable input, while a known numeric zero remains a measured value. If a selected group has no completed orders, AOV and coverage are unavailable. If all selected durations are missing but orders exist, the sales measures can still be calculated while the observed delivery mean is unavailable. One incomplete field should not silently remove otherwise valid sales records.
Review interaction sequences
Change period and channel together, visit an empty selection and use reset. Ensure old bars and explanations disappear or update. Compare May/Web against April/Web, not April/All. Confirm that April’s absent March comparison stays unavailable. In a disposable copy, change an input value and verify that dependent calculations update or a source-validation check fails clearly. Do not call a cached table recalculated simply because it looks plausible.
Check the reading experience
Use labels and numerical alternatives so a reader can understand the chart without relying on colour. Keep controls keyboard-operable with visible focus and a logical order. Use real table headings and captions. At a narrow width, stack controls and charts while keeping the overall page inside its viewport. A wide table may scroll within its own container, provided the content stays reachable.
These checks address specific usability and accessibility concerns; they do not establish complete conformance with an accessibility standard. Record what was actually checked and what remains, such as a fuller assistive-technology review. Avoid placing an “accessible” badge on a prototype merely because its colours look calm. A claim should be supported by an explicit scope of evaluation.
Make freshness honest
The worked example embeds a static fictional fixture. Its edition identifies the supplied material, not the last time a production feed succeeded. There is no server polling, live connector, scheduled update or user-upload process. A real handover would document the source owner, intended refresh schedule, last successful ingestion, validation failures and response process. Do not display today’s clock time as “last refreshed” if no data refresh occurred.
The final review should leave a concise record: test state, expected result, observed result, pass/fail and unresolved issue. Keep the source, metric definitions and editable dashboard with that record. The next reader should be able to reproduce the important numbers and understand how a future change would be validated.
Practical exercise
Use review_checklist.txt and expected_states.json to review your prototype. Check all six main states, reset, table reconciliation, missing-versus-zero behaviour and one changed input on a disposable copy. Inspect keyboard navigation, text alternatives and narrow-screen reading.
Deliverable: a test record with expected/observed results, a list of unresolved issues and a handover note. State honestly whether the source is a static fixture or a live feed; do not invent a refresh time or claim full accessibility certification.
Open the worked solution
A strong test record includes May/All = €3,420, 20 orders, €171 AOV, 95% coverage; April/All = €2,900, 20, €145, 70%, with no March comparison; combined/All = €6,320, 40, €158, 82.5%, with no prior two-month comparison. May/Web and May/Store must match the selected-channel baselines. Empty selection shows zero count/value but unavailable AOV, coverage and relative change.
For a changed-input test, increase a verified completed order’s value in a disposable copy. Its period/channel value and AOV should update while order count stays unchanged. Changing a missing delivery measurement to a known zero should increase measured count, not missing count. Keep the original fixture intact. The worked example is generated from embedded data; changing a separate CSV beside it does not automatically refresh that HTML.
The handover states: “Static fictional fixture edition 2026-10-15.5. No live ingestion or automatic refresh. Period/channel controls recalculate the embedded data. Source changes require rebuilding or replacing that embedded dataset and repeating reconciliation and interaction checks.” Record limited keyboard/responsive checks as such rather than asserting complete standard conformance.
One more question.
Selected usability references: chart descriptions, meaning beyond colour and keyboard operation. These references do not certify this example.
How these learning materials are prepared and maintained · Report a correction: support@datacareerhub.io
Read and practice without an account. Sign in with an existing account to save lesson completion.
Next lesson →Fictional training data. Portfolio practice is not a claim of client work or an accredited qualification. Explore practical lessons, worked examples and projects. Follow a published path, practise the method and build work you can explain.