Content modelling is the design of the data shapes that editors author against and that the front-end consumes. In a headless CMS (Contentful, Sanity, Sitecore XM Cloud, .NET ABP), every editorial concept -- Article, Author, Service, Hero Block, Page -- is a typed content entry with named fields, validation, and relationships.
Done well, content modelling is the highest-use decision in the whole CMS implementation. Done poorly, it traps the team in a model the front-end fights for the next three years.
The Dcrayons rules: model around editorial workflow (not page layout), reference fields over duplicated content, validation at the model level (not the front-end), deprecation over deletion. The single biggest failure mode is content-type drift: implementation week one has 8 clean types, year one has 34 near-duplicates, and removing any one is too risky.