The idea
A fact table records events at a defined grain; dimensions describe entities such as products and dates. Relationships affect how filters propagate. Establish unique dimension keys and avoid joining tables without understanding cardinality, because repeated matches can distort reported results.
Worked example
A fictional sales fact contains one row per order item and a product key. The product dimension has one row per product. Duplicating a product key in that dimension undermines the intended many-to-one relationship and deserves a source correction.
Try it
Draw a sales fact and product and date dimensions. State the grain and keys. Include a duplicated dimension key and explain why it must be investigated. Mark which table a product category filter should use.