
Measurement
Part of Reputation management measurement in review software
Reputation management reporting dashboard: reconciliation first
Reputation management reporting dashboard design for England teams, covering source reconciliation, missing data, response queues and export checks.
A reputation management reporting dashboard should let its reader trace a change to the records behind it. Build the first version around one England branch population and a fixed reporting interval. Agree the extraction cut-off with the people supplying review and operational data before choosing charts or colours.
This workflow specifies a dashboard design. It does not describe a tested system, an actual business dataset or measured improvements.
Sources to reconcile might include Google Business Profile, Trustpilot, Reviews.io, Feefo, TripAdvisor and Facebook. Build the dashboard in a business-intelligence tool such as Looker Studio, Power BI or Tableau, and keep the raw extract in storage you control.
What to take away
- Build a reconciliation table first to map records and show unmapped counts before calculating any headline score.
- Separate panels for reviews, responses, interactions and outcomes so each answers a distinct question.
- Make empty, late and provisional data visible with dated warnings and revision notes.
- Write a short interpretation beside each change, distinguishing recorded movement from possible explanation and evidence needed.
- Check exports by recalculating a rate and following a late record through revision before circulation.
Start with a reconciliation table
Import the platform and branch identifiers as text, preserving leading characters and the original source labels. Keep the raw extract unchanged in controlled storage and work from a versioned copy. Record when the extract was requested, when it completed and which period it covers.
Reconciliation before scoring
- Import identifiers as text
- Keep raw extract unchanged
- Map to approved branch list
- Quarantine unmatched identifiers
- Show unmapped count
- Build review movement table
Map each record to an approved branch list. Quarantine unmatched identifiers for investigation instead of assigning them to the nearest-looking name. Show the unmapped count on the dashboard because excluded branches can change the apparent result.
Build the review movement table before calculating a headline score. Its columns should distinguish newly observed reviews, amendments, disappearances and the closing visible stock. Where a platform cannot explain a disappearance, use an unresolved status. Avoid giving it an invented moderation reason. A workable schema uses these fields: source, platform review ID, branch identifier, branch name, review date, rating, response status, response date, last seen date, movement status (new, amended, disappeared, visible) and unmapped reason. Match on the platform review ID plus the branch identifier, and keep the unmapped count visible.
Arrange panels around different questions
The review panel should display the period's additions and the captured rating distribution by platform. Put the relevant observation date beside each stock figure. A current total and a historical monthly flow should never share an unqualified total label.
Panel questions and contents
Panel
- Review
- Additions and rating distribution
- Response
- Cohort, replied, waiting, timing
- Interaction
- Profile activity with coverage label
- Outcome
- Completion definition records
Must show
- Review
- Unqualified total label
- Response
- Let quickest cases explain queue
- Interaction
- Present combined source as organic
- Outcome
- Mix attribution into outcome view
Must not do
- Review
- Response
- Interaction
- Outcome
The response panel needs the eligible cohort, replied count, waiting count and the timing convention. Give the unanswered age distribution enough space to remain visible when managers filter by branch. The quickest answered cases cannot explain the whole queue.
An interaction panel can contain profile activity. Google's performance documentation says its performance information includes organic results and Google Ads. Label that coverage explicitly when importing it. Do not present a combined source as organic-only activity.
Use a separate outcome panel for records that meet the business's stated completion definition — for example, a booking, a sale or a resolved case. Put attribution credit in another view, naming the model (such as last touch) and stating the observation window in the heading. A side-by-side chart may prompt investigation, but its alignment does not establish why an outcome changed.
Make empty and late data visible
Distinguish a recorded zero from an unavailable field, a failed refresh and a suppressed result. A failed connection should leave a dated warning, not a green status inherited from yesterday. Retain the last successful capture time even when the system cannot fetch a new one.
Empty and late data states
State
- Recorded zero
- Distinct from unavailable
- Unavailable field
- Labelled as unavailable
- Failed refresh
- Dated warning, not green
- Suppressed result
- Marked suppressed
- Provisional period
- Shown differently from closed
- Late arrival
- Revision note maintained
How to show it
- Recorded zero
- Unavailable field
- Failed refresh
- Suppressed result
- Provisional period
- Late arrival
Show provisional periods differently from closed periods. Decide how late arrivals will revise a previous report and maintain a revision note. If readers download a table, include its version and extraction date so it remains interpretable outside the dashboard.
For shared accounts covering England, Scotland, Wales and Northern Ireland, use an explicit geography filter with a value for each nation. The England branch population is the approved list of branches used as the match target; records from the other nations should be filtered separately. Label records that cannot be allocated. A business address may support a branch classification, but it cannot establish the residence of every reviewer or customer.
Write a short interpretation beside each change
Ask the report owner to distinguish a recorded movement, a possible explanation and the evidence needed to investigate it. HM Treasury's Magenta Book treats monitoring and evaluation as related but distinct activities. A dashboard can reveal where investigation is needed without becoming an impact evaluation. For review handling, the Competition and Markets Authority enforces the Digital Markets, Competition and Consumers Act 2024, which addresses fake and suppressed reviews.
Annotating a change
Is the movement recorded in the data?
State the recorded movement
Do not invent a commercial story
Add annotations for collection changes, branch openings and altered review-request procedures where those events are documented. Do not fill gaps with a plausible commercial story. Let users inspect the affected records and the definition in force at that point.
Check the export before circulation
Recalculate one rate from its displayed numerator and denominator. Follow one late record through the revision process, then verify that the downloaded table preserves missing-value labels and source dates. Confirm that every filter changes both parts of each rate consistently. Release the report only when its owner can explain those checks and any unresolved exceptions in ordinary language.
Before you act
- Agree extraction cut-off with data suppliers before choosing charts.
- Import identifiers as text and keep raw extract unchanged.
- Quarantine unmatched identifiers for investigation.
- Label interaction panel coverage explicitly when importing.
- Distinguish recorded zero from unavailable, failed or suppressed data.
- Recalculate one rate from displayed numerator and denominator.
Common questions
What should a reputation dashboard start with?
It should start with a reconciliation table. Import platform and branch identifiers as text, map records to an approved branch list, and quarantine unmatched identifiers. Show the unmapped count on the dashboard because excluded branches can change the apparent result. Build the review movement table before calculating a headline score.
How should empty and late data be handled?
Distinguish a recorded zero from an unavailable field, a failed refresh and a suppressed result. A failed connection should leave a dated warning, not a green status inherited from yesterday. Show provisional periods differently from closed periods. Decide how late arrivals revise a previous report and maintain a revision note.
What checks are needed before circulating the export?
Recalculate one rate from its displayed numerator and denominator. Follow one late record through the revision process. Verify that the downloaded table preserves missing-value labels and source dates. Confirm that every filter changes both parts of each rate consistently. Release only when the owner can explain those checks and unresolved exceptions.



