Retail media dashboards should show coverage gaps and estimates as plainly as results

A retail media reporting dashboard should reveal what its numbers count before asking an England retailer to act on them. Build it around a defined property and advertising format, with separate displays for delivery observations, attributed sales and analytical estimates. The arrangement here is a proposed design method, not a tested dashboard.

Research completed for this page on 6 September 2026 used public guidance only. No account, campaign export or shopper dataset was accessed. All values and performance comparisons remain unpopulated until the retailer provides valid evidence.

Establish the report context

Name the retailer-controlled inventory, campaign scope and intended decision. Set the report period and time zone explicitly. Include extraction time and dataset version so an exported chart can be interpreted after the underlying data changes.

Require every tile to carry or reference its event, unit, denominator, cohort, period, time zone, attribution window, exclusions, corrections, source and uncertainty. Counts use a not-applicable denominator; non-attributed measures use a not-applicable attribution window. Missing definitions prevent a tile from presenting a numerical result.

Avoid a global filter that silently changes one tile's population while leaving another unchanged. This is a design question to inspect in the eventual dashboard, not behaviour observed in a named product.

Separate delivery states

Create distinct count fields for ad requests, served or returned responses and rendered placements. Each needs its own source-event definition. Do not imply a one-to-one relationship or fill missing stages through an assumed conversion factor.

Use the IAB/MRC January 2024 guidelines as an international reference when specifying media measures. A viewable-impression tile should identify the adopted criteria and coverage; an attention tile requires separate evidence and otherwise remains UNKNOWN.

If displaying a click-through ratio, expose the qualifying-click numerator and eligible impression denominator. Both must share the stated cohort and period. Keep empty or undefined denominators visible as unresolved calculation states.

Give sales interpretations their own space

A recorded sale and an attributed sale need different labels. For credited sales, show the interaction rule, lookback window, unit and treatment of returns. Prevent a reader from mistaking a count of orders for a count of items.

Put modelled values beside their assumptions and uncertainty, with a clear distinction from observed records. A causal-incrementality display requires its own evaluation design and outcome definition. Ordinary attribution does not supply that justification.

The proposed dashboard should allow those unavailable measures to remain absent with an explanation. A blank field is more informative than an invented nil result that appears to represent completed measurement.

Explain gaps in observation

The ICO's advertising storage and access guidance covers consent in the relevant online-advertising context. Have qualified privacy staff assess actual collection and processing; dashboard convenience does not determine permission.

Document the population covered by the available data and any known exclusions. Do not assume unobserved shoppers behave like the measured cohort. If an estimate extends beyond that cohort, expose the model and its limits rather than merging it invisibly into a total.

Preserve corrections and comparison limits

Attach an explanation when a revised export changes a tile. Record the affected period, definition and source version, and retain the earlier issue. This design imposes no invented supplier refresh schedule or response deadline.

Before showing period comparisons, establish consistent scope and counting rules. A metric-definition change should be disclosed as such, without automatically describing its effect as better or worse campaign performance.

Check the actual dashboard before use

Ask a real reviewer to trace a displayed value back to its event source and definition record. Record what was inspected and what remains untested. This page makes no usability, accuracy or launch verdict; publication awaits genuine authorship, qualified review and disclosure of publisher relationships.