What do economists mean by marginal analysis?

Marginal analysis is often presented as one of the core tools in economics, usually summarized by the idea that decisions are made by comparing marginal benefits and marginal costs and continuing an activity until MB = MC.

My question is about how this framework should be understood outside of textbook settings.

In theory, marginal analysis explains a wide range of behavior: firms hiring workers, consumers choosing quantities, and governments expanding or contracting programs. But in practice, many decisions seem discrete rather than marginal, and people rarely have clear information about marginal costs or benefits.

So how do economists interpret marginal analysis in applied contexts?

• Is it mainly a predictive model rather than a literal description of how agents think?
• In what types of markets or policy settings does marginal analysis work best?
• Where does it break down due to uncertainty, behavioral constraints, or institutional rules?

I’m especially interested in how economists reconcile the centrality of marginal analysis with real-world frictions and non-marginal choices.