We gather the questions
What question the report must answer and who will look at it.
We start from what the management wants to know: the list of reports first, and only then the data model to support it.
What question the report must answer and who will look at it.
The accounting system, the website, the CRM, spreadsheets. We decide what comes from where.
A single model where products and counterparties are matched across the different systems.
The data is pulled on a schedule that does not disturb people working during the day.
The data for each report is calculated in advance, so it opens without delay.
Dashboards with charts or an export into the familiar spreadsheet format.
The first dataset appears in a month to six weeks. The next ones come faster, as the foundation is already there.
We start with the two reports that take the most time to assemble by hand.
The structure, mapping the reference data between systems, the refresh schedule.
The load, the dataset and a reconciliation of the result against the manual calculation.
New datasets as requests come in, on the foundation already built.
The first test of a data warehouse is matching the manual report. If the figures differ by so much as a penny there will be no trust: staff will quietly carry on with their own files. So a reconciliation against the old calculation is always done, and any discrepancy is explained before go-live.
Built-in reports show only that system's own data. As soon as you need to combine sales with site traffic and enquiries from the CRM, the standard tools fall short. Heavy queries against the live database also slow everyone down.
At the start you can do without one. Ready datasets are easy to deliver as spreadsheets: people know them and no training is needed. A specialised product is required where a wide circle of staff look at the reports and each needs to change the slices themselves.
Management reports usually need one refresh overnight. Continuous streaming is rarely necessary, costs more and puts extra load on the live databases. It is sensible to start with a daily cycle and add frequency only where it is genuinely needed.
List the reports currently assembled by hand and the source systems. We will build the first dataset and reconcile it against your own calculation.
Request received
It is already with a manager. You will get an answer within the working day, and urgent requests go to the duty engineer immediately.
There is no such city in the list. Check the spelling.