Cost estimates are classified by how accurate they are, which in practice tracks how much effort goes into them and which project phase they’re produced for. Three standard categories:
Order-of-magnitude estimate. Accuracy roughly ±30-50%. Produced very early — when the project is still a sketch. Based on scaling rules, historical analogues, and quick parametric models. Cheap to produce, used to screen ideas: is this project even worth a closer look? Conceptual designs typically have order-of-magnitude estimates only.
Semi-detailed / budgetary estimate. Accuracy roughly ±10-20%. Produced when the project has a defined scope, a preliminary design, and a more detailed work breakdown. Used to set budgets, request capital approval, and decide between competing project alternatives. The most common estimate class for present-worth and IRR comparison.
Definitive / detailed estimate. Accuracy ±5-10% (sometimes tighter). Produced when the design is mostly complete, vendors are quoting actual prices, and the work breakdown structure is detailed. Used for final approval, vendor contracts, and execution control. Expensive to produce — usually only done for projects that are going forward.
The general principle: estimate accuracy increases with project maturity, and so does the cost of producing the estimate. Early-phase estimates are cheap and crude; late-phase estimates are expensive and precise. The work breakdown structure (WBS) — how finely the project is decomposed into discrete tasks — is shallower for early estimates and deeper for detailed ones.
A useful heuristic: an order-of-magnitude estimate that’s already five times the budget envelope is enough to reject the project. There’s no point producing a detailed estimate of a project you can already tell is unaffordable.
For specific techniques used at each level, see Cost index, Unit cost estimation, Parametric cost estimation, Power-sizing model, and Learning curve model.