Reporting Architecture: One Data Model, Many Reports
When the sales report and the finance report disagree, meetings turn into arguments about numbers. A shared data model means every report uses the same definitions, so the numbers reconcile.

The sales director’s report says last month’s revenue was JOD 1.42 million. The finance report says JOD 1.37 million. The meeting spends twenty minutes arguing about which number is right, and nobody discusses what to do about the result. Both reports are probably correct by their own logic. The real problem is that each was built separately, with its own rules, from its own extract.
Why separate reports drift apart
In most mid-sized organisations, reports grow one at a time. Someone needs a sales view, so an analyst builds a workbook from an enterprise resource planning (ERP) export. Finance builds its own pack. Operations has a dashboard made by a consultant three years ago. Each one defines revenue, margin or “active customer” slightly differently, applies its own filters and refreshes on its own schedule.
The differences are often small and reasonable. One report counts invoices, another counts orders. One excludes returns, another nets them off later. One uses invoice date, another delivery date. But when nobody can explain the gap in a meeting, trust in all the reports falls, and managers go back to asking someone to “just pull the numbers” by hand.
There is a cost in time as well. Every separate report has its own extract, its own cleaning steps and its own maintenance. When a product category is renamed or a new branch opens, someone has to update five workbooks instead of one, and usually one of them gets missed.
The idea: one model, many reports
A better reporting architecture separates the data and its business rules from the reports that display it. Data is collected and cleaned once, business definitions are written once in a shared model, and every report reads from that model. The layers look like this:
| Layer | Purpose | Typical owner |
|---|---|---|
| Sources | ERP, point-of-sale (POS), customer relationship management (CRM) and human resources (HR) systems, plus spreadsheets that hold the original data | System owners in each department |
| Staging and cleaning | Copies of source data, standardised and checked | Data or business intelligence (BI) lead |
| Shared data model | Business tables and agreed measures such as net revenue, gross margin and on-time delivery | BI lead, with finance and operations sign-off |
| Reports and dashboards | Views for specific audiences and decisions | Report authors in each team |
In a Microsoft environment, this might be a central Power BI semantic model (the shared dataset holding tables and calculations) that many reports connect to, fed from a small data warehouse or a set of cleaned tables. Other BI tools follow the same principle under different names. The important part is the separation, not the product.
A scenario: a trading group with two companies
Imagine a family-owned group in Amman with a distribution company on one ERP and a retail company on an older accounting system. The owners want a monthly group view, and each company’s managers want their own operational reports.
Without a shared model, someone exports from both systems every month, maps accounts and products by hand, and builds a group pack in Excel. Each company also has its own reports with different definitions. When the owners ask why the group margin differs from the companies’ figures, the answer takes days.
With one model, data from both systems is loaded nightly into a staging area. Product categories, customer types and branches are mapped to one shared list. Net revenue and gross margin are defined once, with finance agreeing the rules. The group pack, the distribution sales dashboard and the retail branch report all read from the same model. They show different views, but the numbers reconcile because they come from the same definitions.
Design principles that keep the model useful
- Define measures in writing first. Before building, write a short definition for each key measure: what it includes, what it excludes and which date it uses. Finance and operations should both sign it off.
- Build around business entities. Customers, products, branches, dates and transactions are the backbone. Get these shared lists right and most reports become simple.
- Keep business logic out of individual reports. If a report author needs a new calculation, add it to the model so everyone benefits and nobody creates a competing version.
- Control changes. Changes to shared measures need a review, because they affect every report at once. A simple change log is usually enough.
- Apply security once. Rules on which rows of data each person may see, such as branch managers seeing only their branch, belong in the model, not in copies of reports.
- Show freshness. Every report should display when its data was last refreshed.
Getting there one subject at a time
- List existing reports. List the reports people actually use, who uses them and which measures they show. You will usually find that a few reports carry most of the weight.
- Pick the measures that cause arguments. Revenue, margin and stock value are common starting points.
- Build the first version around one subject area, usually sales, and move the most-used reports onto it.
- Retire the old versions. Once a report has moved, archive the old workbook so people stop using it.
- Extend one subject at a time, adding stock, purchasing or finance as the model proves itself.
Expect the definition work to take longer than the technical build. Agreeing what “net revenue” means across sales and finance can take several conversations. That time is well spent, because every report built afterwards inherits the agreement.
Reports that agree with each other
A shared model turns reporting from a collection of personal workbooks into a reliable service. Meetings move from arguing about the number to discussing what it means. Designing this kind of data architecture, with definitions agreed by the business, is central to our Enterprise Systems service, and it creates the foundation for automated reporting later. If you want a starting point, put your current sales and finance revenue figures side by side for one month and write down every reason they differ.



