Most engineering models come from measurement, not derivation - even when the underlying physics is well understood. System Identification for Engineers teaches the small set of methods that cover the large majority of real identification jobs: designing an experiment that genuinely excites the dynamics you care about, cleaning and preprocessing real sensor data, fitting first- and second-order models by hand and by code, least squares and its recursive (online) form, ARX/ARMAX/Output-Error model structures, subspace state-space identification, closed-loop and nonlinear identification, and validating a model honestly before trusting it in deployment. Every method is built plain-language-first, worked through with runnable Python (NumPy, SciPy, python-control) on clearly-labeled synthetic datasets, and tied together in three full case studies that run the entire pipeline start to finish: a DC motor identified from encoder data, a quadrotor attitude axis identified under closed-loop and mild nonlinear conditions, and a battery equivalent-circuit model tracked across simulated aging. Written for working controls, robotics, and mechatronics engineers who need a model that holds up in deployment - not a research paper.