Data & Analytics
Reporting, analytics, data warehousing, and business intelligence.
Workflow
- 01Data Collection
- 02Data Cleansing
- 03Data Analysis
- 04Data Reporting
- 05Decision Making Support
If you change this
It implies 4 topics across one module.
See the path ahead →Why it matters
Every other module produces data. This one decides what it means. Without ownership and definitions, an organisation ends up with several versions of its own performance and an argument at every meeting.
- Definitions are the product: A metric without an agreed calculation is a source of dispute.
- Trust is binary: Once a report is caught being wrong, it stops being used regardless of later fixes.
- It outlives projects: Data investment pays back across programmes; it is also the first thing cut.
The workflow
- 01
Source contract
Owner, meaning, frequency and quality promise agreed for each source.
- 02
Ingestion
Data received on a controlled cycle, with failures visible.
- 03
Quality rules
Completeness and validity tested, with failures going to a worked queue.
- 04
Modelling
A shared representation the business recognises.
- 05
Metric definition
Calculation and meaning agreed and recorded.
- 06
Publication
Presented where the decision happens, with the as-at date visible.
Put the as-at date on every report. Most disputes about whether a number is wrong turn out to be disputes about when it was taken.
What you need in place
- Domain ownership: A named owner per data domain, accountable for quality.
- A metric register: Definition, calculation and owner, visible to anyone using the number.
- A worked exception queue: Quality failures that someone actually resolves.
- Reconciliation to source: Particularly finance, where the ledger is the arbiter.
- Access control: Especially for person and customer data.
Questions to ask
01Ownership
Who owns each data domain?
If the answer is IT, nobody owns meaning.
02Definitions
Where is the metric register, and who maintains it?
Without one, every report is a new negotiation.
03Reconciliation
Does reporting reconcile to the ledger?
If not, two numbers circulate and one is wrong.
04Quality
What happens when a quality rule fails?
A queue nobody works is a rule that does not exist.
05Timeliness
How stale is the data when decisions are made on it?
Daily reporting on a weekly feed is a trap.
Metrics and KPIs to track
Quality rule pass rate
By domain and source.
Exception queue age
How long failures wait to be resolved.
Metric register coverage
Reported measures with an agreed definition and owner.
Reconciliation variance
Reporting against the ledger.
Report usage
Which reports are actually opened.
Covers