All models are wrong, but some are useful.

All measurements are a representation of reality, some more accurate and more precise than others. In many instances we become accustomed to acting as if the measures are reality but they are not, as reality periodically reminds us. Likewise with models with allow us to produce forecasts of greater or lesser accuracy and precision. An idea which Box and Draper neatly capture.

From Empirical Model-Building and Response Surfaces by George E. P. Box and Norman R. Draper, p. 424

All models are wrong, but some are useful.