Five acceptance tests for a PR agency's reporting dashboard

A PR agency's reporting dashboard fails quietly. A header goes stale, a percentage loses its denominator, or a drill-down opens for the wrong person. Five acceptance tests let a buyer in England catch these faults before relying on the screen an agency supplies. Run them on the blank template first, then again on real records.

What to take away

Test 1: the header and the refresh rule

The header passes if a reader can judge how fresh the figures are without asking anyone. It must show timezone, last successful refresh, source versions and open corrections. If any of the four is missing, the dashboard fails this test.

Next, check the reporting period against the buyer's calendar. Typical UK reporting periods follow the calendar month, the quarter or the financial year, which for many UK organisations runs from 6 April to 5 April. Match the refresh to the fastest decision the dashboard serves, from daily for coverage counts to monthly for reconciled measures.

Build the blank template in a spreadsheet or in a tool such as Microsoft Power BI, Google Looker Studio or Tableau, and run the same header test in each. The tool changes automation and export options, not what counts as evidence. A header that passes in a spreadsheet must still pass after a move.

Test 2: each panel with empty data

A panel passes if it can sit empty without misleading anyone. Load the template with no values and read every panel as a stranger would. A blank panel should say what is missing, not show zero, green or a default percentage.

Apply the HOLD rule next. If a rate lacks a reconstructable denominator, show HOLD instead of a percentage. Every card should also link to its metric dictionary, so a reader can rebuild any rate from its numerator and denominator.

Colour is a separate test. It must not quietly turn an arbitrary threshold into evidence, so thresholds stay blank. Targets require a defensible baseline, consequence analysis and named approval. A change may reflect seasonality, media events, platform processing, missing records or a revised query rather than agency activity.

Independent editorial coverage belongs in an observed media panel, not in Delivery. A newsroom's decision to run a story is not agency-controlled delivery, and potential reach stays separate from observed attention.

Test 3: who can see what, and what is missing

Permissions pass when each reader sees only the detail their decision needs. Use aggregated or suppressed views where personal detail is unnecessary. Restrict drill-down access and give it a documented purpose, retention period and review owner.

Tracking is not neutral plumbing. The ICO's finalised guidance on storage and access technologies follows its earlier cookies guidance and reflects the PECR changes made by the Data (Use and Access) Act. It does not approve the buyer's implementation.

Then test the loss. The dashboard should state the collection route, permission state, excluded population and known loss instead of filling gaps with silent estimates. Ask the decision owner to confirm each of the four in writing.

Test 4: the audit trail

The audit trail passes if a changed value can be traced back to its original. When a value changes after publication, retain the original, replacement, reason, authoriser and reports affected.

The NCSC's cloud audit-information principle supports checking what logs a service supplies, their format, access and retention. It is security guidance, not proof that dashboard figures are accurate. Keep a separate record of source refreshes, query edits, permissions, exports and corrections.

Run one deliberate change as the test. Edit a dummy value, then try to rebuild the previous version from the record alone. Pause the panel if the source version is unknown, the query cannot be reproduced or access evidence is missing.

Test 5: the exported report

An exported file passes only if it is accessible in its own right. An accessible screen does not guarantee an accessible PDF or spreadsheet, so export a real report and test that file, not the screen it came from.

Government inclusive communication guidance is aimed at government work, but its prompts on formats and channel choice help form test tasks. Check heading order, table structure, text alternatives, reading order, contrast, keyboard use and plain-language notes on real exports with a named reviewer.

The last page of each report should state the next decision, owner and review date. If a panel cannot change a decision, remove or redesign it. The dashboard stays on HOLD until the actual data route, permissions, calculations, security controls and accessible outputs have been tested.

Before you sign off

Common questions

What should a panel show when the denominator cannot be rebuilt?

It should show HOLD instead of a percentage. A rate that lacks a reconstructable denominator cannot be checked, so the panel stays on hold until the numerator and denominator can be rebuilt from records.

Does the ICO guidance approve our tracking set-up?

No. The ICO's guidance does not approve the buyer's implementation. The dashboard should state the collection route, permission state, excluded population and known loss instead of filling gaps with silent estimates.