Building an accounting dashboard in three layers, with uncertainty shown honestly, and what to ask your accountant for

An accounting dashboard is useful when it connects the evidence behind a measure to a decision someone is authorised to make. Build around that route, rather than starting with charts or whatever fields the software happens to expose. Show uncertainty where the figures appear.

What should the dashboard help people decide?

It should help a named person decide whether to act, wait, investigate or correct the underlying record. A dashboard earns its place when its measures and evidence support that choice, rather than simply displaying a large collection of charts.

Start by identifying the recurring decision and the person responsible for it. Operational staff may need to respond to exceptions, managers may need to understand patterns, and clients may need to see delivery status. The decision should shape the view, rather than the available fields dictating what appears.

Who should see each view?

Give each person only the view needed for their task, with access determined by the decision they can make. Operational staff need exceptions, managers need patterns, and clients need delivery status; one shared screen need not expose every underlying field.

An engagement owner may need threatened deadlines and missing client inputs, while a quality lead may focus on serious corrections and recurring causes. Write down each user's question and permitted response first. The same record can support different views without making every piece of information visible to everyone.

Keep separate views where their purpose or access rules differ. A client status can show whether a deliverable is progressing without exposing internal personnel information or customer financial details. Filtering a report does not, by itself, make this separation happen; it needs to be part of the design.

How should a measure be defined?

Define every measure before selecting its visual form. Record what is counted, the population and exclusions, the time basis, source, refresh timing, owner, warning point and expected response, then keep the definition versioned and its explanation visible beside the displayed value.

Use a short measure record with these four checks:

For example, for VAT filing timeliness, the numerator is returns filed by the HMRC deadline and the denominator is returns due in the period. The workflow system is the source and the VAT compliance lead owns the warning; it turns amber when a return due within seven days has not been filed.

The Government Digital Service material on performance data shows how a service objective can guide measure selection and improvement. Choose a measure because it answers a practical question, not merely because data is available. Keep the brief explanation visible and retain the full method in the register.

What belongs in each layer?

Use three layers. Layer 1 shows what needs attention now. Layer 2 explains how work and quality change over time. Layer 3 lets an authorised reviewer reach relevant source records. This order keeps urgent action visible while limiting detailed evidence to people permitted to inspect it.

The first layer should surface filing deadlines at risk, including HMRC VAT returns and Companies House accounts, as well as rejected submissions, high-consequence errors, severe incidents and exceptions without an owner. Its purpose is to show what needs attention, not to explain every cause.

The middle layer explains whether work is flowing, where it is ageing and which service or quality patterns need attention. Choose service metrics to support those decisions. Typical layer 2 views are:

The detailed layer should offer a clear route to source records without collecting or displaying unnecessary personal information.

How can I verify the figures?

Trust depends on knowing where a figure came from and confirming that it still means the same thing. Map the source, extraction, transformation and calculation, then reconcile totals against originating systems on a schedule before releasing the view.

Check for missing refreshes, changed field meanings and duplicate records as part of that review. Retain record exports and use role-based access, so reviewers can trace a result while access to its underlying information remains limited to the appropriate people.

Freeze a period after review and label any restatement so later changes do not silently rewrite an earlier result. The ICO's data security guidance describes organisational and technical measures proportionate to risks involving personal information. Restricting access also supports data protection principles.

How should uncertainty be displayed?

Show the period and count beside a rate, and identify small samples, estimates, incomplete feeds and changed definitions. Put uncertainty next to the value it affects, so readers can distinguish a firm result from one that needs qualification.

Keep client waiting time separate from time spent waiting inside the practice, while retaining end-to-end elapsed time to show the whole delay. Place benchmark estimates, missing feeds and revised definitions beside affected values, rather than presenting them as settled results.

Do not reduce every result to one red, amber or green score. A favourable overall picture can hide a late statutory submission or a confidential-data incident, so important exceptions should remain visible alongside summary measures.

What should a client see on it?

If you are choosing an accountant or bookkeeper, the client-facing layer is the part that matters to you. Its financial content should be plain: income and expenses for the period, cash at bank against what the books show, VAT due and the date it is payable, and unpaid invoices.

It should also show delivery status, such as which return or set of accounts is with the firm, what is waiting on you and the filing deadline. Estimates and unreconciled items should be marked as such, not blended into the totals.

Questions to ask a firm about its dashboard

Ask any firm you are considering three things. Can I see a sample client view with real figures removed? Which numbers are reconciled to bank statements and which are estimates? Who corrects a wrong figure, and how fast?

A firm that cannot answer is not necessarily poor, but it gives you less to compare. Put the answers beside its quality checklist results and the fees it quoted.

How can people use and challenge the results?

Make the display usable without relying on colour or a mouse. Provide plain labels, readable contrast, keyboard operation and tabular downloads suited to the audience, then test accessibility requirements with representative users and assistive technology.

A small practice can use Excel or Google Sheets for the register and simple views, with CSV or PDF exports for evidence. A BI tool, such as Power BI or Looker Studio, can support recurring views once source exports are stable; match the display to the users and evidence.

Assign someone to check refreshes, data quality and anomalies before release. Record each meeting's decision, owner and due date beside its evidence, then review whether the action changed the measure or caused an unwanted side effect. Annotate scope, staffing, software and workflow changes so a shift is not mistaken for improved performance when definitions or workload have changed.

Give people a route to report a wrong total or misleading chart. The owner should log the investigation, amendment and any decisions affected. Retire charts no longer used, investigate gaming and revisit definitions after service changes; the display should reduce uncertainty and prompt proportionate action.

What should I do first?

Name the people, decisions and permitted actions before choosing a chart. Then define the measures, trace their evidence and assign an owner to check information before release.