Role
Business intelligence
How many times a month does somebody ask you for a report that already exists? If it is often, the problem is not the number of dashboards, it is that nobody knows which of them are true.
Intro call
Four questions we would start with
How many times a month are you asked for a report that already exists?
If it is often, the report catalogue is unreadable or people do not trust it. Both causes need a different answer than another dashboard.
What do you not do, even though you know it has to be done?
Usually: you do not remove dead reports and you do not put metric definitions in order. Always because new requests are more urgent.
Who decides how a metric is calculated?
If nobody does, two reports will show two different numbers and both will be correct under their own definition.
How long does it take before a new report exists?
If it is weeks, departments build their own summaries in spreadsheets - and that is the start of a parallel world of data.
We take over the building and the upkeep, you keep the model and the definitions
Gathering requirements, building views, maintaining the catalogue and clearing it out can be taken off the desk. Deciding how a metric is calculated and what it means stays with a person.
- A metric has one definition. - Recorded, with an owner, visible on every report it appears in.
- A dead report is visible. - A dashboard nobody has opened in six months is pointed out - and removing it is sometimes more important than building another.
- A request for a report starts with checking whether it already exists. - Before a new view is built, it is clear what is already there and how it differs.
Scope
What can be taken off a BI team
Nineteen example tasks. Each one has a human approval point. Below is a selection, and the scope we fit to your reporting environment.
Gathering reporting needs
- One route for raising a reporting need
- Pointing out existing reports that answer the question
- Asking which decision the report is to serve
- A signal when a request duplicates an existing view
The model and the metrics
- A catalogue of metrics with definitions and owners
- Pointing out metrics calculated differently in different reports
- Linking a metric to the reports it appears in
- A register of definition changes (approval: the metric owner)
Building dashboards
- A draft view based on the need raised (assessment: the specialist)
- Consistent naming and formats across reports
- Checking the numbers agree with other reports
- Documentation of a report's sources
Rollout to the audience
- The list of recipients and their permissions
- Introductory material for the audience
- A summary of the questions asked after the rollout
Upkeep and clearing out
- Statistics on report opens
- Pointing out reports not opened for a long time
- A signal when a report rests on an out-of-date source
- A proposal to withdraw a report (decision: the specialist)
A day's work
Today
- Requests for reports arrive by email and at meetings
- With each one you start by working out whether something similar exists
- Metric definitions live in the heads of a few people
- Two reports show different numbers and nobody knows why
- You do not know which reports anyone opens
- You put the clearing out off, because new requests are more urgent
With AI assistance
- A request comes in by one route, with the question about the decision to be made
- Existing reports that answer the question are pointed out
- Metric definitions are recorded, with an owner
- A divergence between reports is flagged
- You can see which reports are opened and which are dead
- The clearing out has a list, so it can be done in a few hours
- Readers who help themselves
fewer requests for a report that already exists
it is clear what exists and how it differs
- Time to a new report
a shorter road from the need to the view
the draft comes from the model, not from scratch
- Dead reports
the catalogue is smaller and more credible
reports nobody opens are pointed out
- One source per metric
the same number in every report
the definition has an owner and is recorded
We start with the metric definitions, because trust in everything else depends on them.
- Step 1
A catalogue of metrics with owners
The definition, the owner, the reports the metric appears in.
- Step 2
Detecting divergence
Metrics calculated differently in different reports.
- Step 3
The intake for reporting needs
A request with the question about the decision and existing views pointed out.
- Step 4
Clearing out the catalogue
Open statistics and withdrawing dead reports. Last, because it needs a few months of usage data.
Common questions
Questions that come up at the first meeting
How is this different from a data analyst and a data engineer?
The engineer is accountable for the data existing. The analyst answers questions with it. BI is accountable for the organisation being able to answer for itself - and that is why what matters here is not analysis but a shared definition and trust in the numbers.
Will the BI team be smaller once people help themselves?
Self-service does not reduce BI work, it moves it from making reports to making sure everyone counts the same thing the same way. That work is harder and more necessary. A team that is not producing the fortieth dashboard on request works on the model and the definitions, which is what trust in the numbers across the company rests on.
Does this fall under the AI Act?
Reporting is not a high-risk use. The line runs at dashboards about employees: HR and performance summaries need care and a legal basis, because they easily become a tool for assessing people.
Still have a question that is not here? Book a call
Next step
How many of your reports did anyone open last month?
We will show you what BI work looks like when a metric has one definition and a dead report is visible.