Dashboards and business intelligence

A business intelligence consultancy builds the KPI dashboards and management reports a business runs on, in the tool you already license. Every KPI's definition, formula, source and owner is agreed in writing first, so when a number is questioned the answer is already written down.

Last reviewed

One KPI, two definitions, 422 customers apart

Example data

Active customersOctober 2026 (the 12 months to 31 October)

Sales definition

4,812

Customers with at least one order in the last 12 months.

The gap

422

Customers whose only orders in the window were fully refunded.

Finance definition

4,390

Customers invoiced in the last 12 months, leaving out anyone whose orders were all fully refunded.

Definitions log entry
NameDefinitionFormulaSourceOwner
Active customersA customer with at least one invoiced order in the last 12 months that wasn't fully refunded.Count of distinct customer IDs on invoices dated in the 12 months to the end of the reporting month, excluding invoices whose order was fully refunded.finance.invoices, joined to finance.refunds on order IDFinancial controller

NoteThe sales count stays on the dashboard, renamed “Ordering customers (12 months)”.

Example data: an invented customer base. Both counts cover the 12 months to 31 October, and the 422 between them are customers whose only orders in that window were fully refunded.

What this fixes

Four signs the reporting has stopped settling arguments.

Two right answersSales and finance bring different customer counts to the same meeting, and each is right by its own rules.
A tab for every requestThe dashboard has grown a page for each person who asked, and nobody can say which one the board should read.
Month-end by handThe management pack is rebuilt from exports in a spreadsheet every month, by one person.
The quiet failureThe dashboard is showing last Tuesday's data, and nobody noticed the refresh had stopped.

What you get

The definitions log, model, dashboard and handover you keep.

The definitions logOne line per KPI with its definition, formula, source table and owner, kept in a file you own.
The data modelThe tables behind the dashboard, built at the grain the definitions need.
The KPI dashboard and MI packA dashboard for day-to-day use and a management reporting pack laid out for the monthly meeting, both reading the same measures.
Refresh and ownershipThe refresh runs under an account you control, one named person checks that it ran, and every definition in the log has an owner. In Power BI, the refresh failure email goes to that person.
HandoverA walkthrough with whoever will maintain it, and a one-page map from each figure to its source.

A dashboard is finished when someone other than the person who built it can keep it running.

Dashboard development, stage by stage

From agreed definitions to a dashboard your team runs.

  1. 1

    Scoping

    Start from the decisions the dashboard serves and the KPIs they need, then agree each definition in the log. A disputed KPI, like the active customers count above, gets both versions written down and one chosen, with the other renamed.

  2. 2

    Build

    Each measure is written from its line in the definitions log and carries that line's name. Pages come last, in Power BI or Data Studio (formerly Looker Studio).

  3. 3

    Handover

    The person who'll maintain it changes a figure and republishes while the builder watches.

  4. 4

    Afterwards

    The definitions log, the model and the refresh account sit in your systems, under your logins. A definition changes only when its named owner approves it.

Where data visualisation goes wrong

Ownerless KPIs and silent refresh failures, and the guard for each.

A chart for every requestEach stakeholder asks for a chart and gets one, until the dashboard answers no question in particular. The guard is that every page starts from a decision written down at scoping. A request with no decision behind it waits for the next round.
A KPI with no ownerTwo teams disagree, nobody has the authority to settle it, and both numbers live on in different reports. The guard is an owner role against every definition in the log, and only that owner can change it.
A dashboard nobody can refreshThe refresh ran on the builder's login, it fails after they leave, and the dashboard keeps showing old data with no warning. The guard is a refresh that runs under an account you control, with a named person who checks it. In Power BI, the failure email goes to them. Every page also shows when its data was last refreshed.

Definitions get agreed before anything is built. A chart drawn on a number nobody agreed only moves the argument to a better-looking page.

MI reporting and board packs

The monthly pack, built on the dashboard's measures.

MI reporting uses the same measures as the dashboard, laid out for the monthly meeting: fewer figures, each set against budget and last year, with a line of commentary from the person who owns it. The finance dashboard in our KPI library shows that pack on example data, with the formula behind each line.

Power BI

Power BI has a page of its own, Power BI consultant, because star schema modelling, DAX measures and the choice between Pro, Premium Per User and Fabric capacity need more room than a paragraph.

KPI dashboards by industry

Where each sector's KPI arguments usually start.

EcommerceThe shop platform and the finance system count a sale differently.
Subscription mediaA subscriber count depends on the day a cancelled subscription stops counting, and that one rule moves churn.
Financial servicesA loss ratio on written premium and one on earned premium can both reach the board pack, so the log records which one the board reads.

Questions people ask

Choosing a BI tool, and working without a warehouse.

Do you build KPI dashboards for ecommerce?

Yes. The usual set covers net revenue after refunds, orders, average order value, repeat rate and active customers. Each one goes in the definitions log first, with a note of whether its count comes from the shop platform or the finance system.

Which BI tool should we use?

Usually the one you already license, provided someone on your team can maintain it. A tool nobody in-house knows turns every change into a request to whoever built the dashboard. We build in Power BI and Data Studio. If your data sits in a warehouse such as BigQuery, Data Studio connects to it directly.

Can you work with our existing data warehouse?

Yes. The dashboards read from your warehouse through an account you create, and any modelling happens in views or tables inside it that you can see and keep.

What if we don't have a data warehouse?

A small setup can model straight from the source systems' own connectors or from scheduled exports. Once several systems need joining, or history needs keeping that the sources overwrite, a warehouse comes first, and that's data infrastructure work.

Who maintains the dashboards afterwards?

Your team. Handover leaves an owner named against every definition and one person trained to change a figure, republish and check the refresh ran. In Power BI, the refresh failure email goes to that person.

Book a call about your reporting

Bring the KPI that causes the most argument. The first call works out what its definition should be and who should own it.

Book a call