A data model is the description of which tables you have and how they connect. Customers have orders; orders contain products; products belong to categories and have costs; stores sit in regions. Once those links are described, a question such as “margin by category by region” can be asked directly, because the tool knows which tables to combine and on what.
The model is what turns a pile of connections into something people can use. A manufacturer connects its production system, its quality system and its ERP. Each is useful alone; together, related on order number and part code, they answer whether the scrap on line 3 is costing more than the overtime on line 4. Without a model, someone rebuilds those joins in a spreadsheet every week, and each rebuild differs a little.
Data models do not need to be grand. The useful ones are small and specific: ten tables and a dozen relationships cover most businesses. The rule of thumb is to model the questions you actually ask, name things the way the business names them rather than the way the source system abbreviates them, and define shared calculations in the model so every dashboard inherits them. A model nobody but its author understands will be bypassed.
In Klayara, a data model is built by pointing at tables and drawing the relationships between them, with permissions and calculated fields attached at the same level so they follow the data into every dashboard. Read about formulas and permissions.