Nearly every business question has one form: a measure, sliced by context. "Gross sales by genre group and region for the second quarter" sums a number and groups it by descriptions. A normalized schema (LAMP Stack Development) scatters those descriptions across tables and leaves derived ones, such as region or quarter, to every query. A dimensional model splits data the way the question does: facts are the numeric measurements of a business event (quantity, gross amount), and dimensions are the descriptive context you filter and group by (date, book, customer).
The Kimball Group's four-step design process orders the decisions: select the business process, declare the grain, identify the dimensions, identify the facts. The grain comes first because it fixes which dimensions and facts can belong to a row at all.