Moving from pivot tables to a proper data model
The step from a stack of pivot tables to a small star schema is less technical than it sounds, and it changes what people are able to ask.
The ceiling
Pivot tables scale fine until a question needs two sources at once. Sales by category is easy. Sales by category against stock cover, by region, for the same period, is where the copy-paste starts — and every copy-paste is a version of the truth that will drift from the others by next month.
The model
One fact table of movements. Dimensions for product, supplier, store and date. That is the whole thing — four dimensions and one fact, which is far less impressive than the phrase "data model" suggests.
What it buys you is that relationships are declared once. A question combining product category and region no longer requires anyone to align two exports by hand; it is a matter of dragging two fields. Two analysts asking the same question get the same answer because they are reading the same joins.
The date dimension earns its place immediately. Every reporting period question — same month last year, quarter to date, rolling twelve — stops being a formula and becomes a filter.
What it costs
Discipline about master data, mostly. A model exposes every inconsistency the pivot tables were papering over: the duplicate SKUs, the products in two categories, the supplier under three legal names. Those have to be fixed before the model is trustworthy, and fixing them is unglamorous work with no visible output for several weeks.
It is still the right trade. The pivot table stack hid the same problems; it just let you keep publishing on top of them.