Precision is the proportion of positive predictions that are actually positive. It is one way to judge whether a prediction system performs well enough for its intended use, but it should not be assessed alone. section 8.6
Make performance measurable before development
For proposed customer-churn and demand-prediction systems, include precision and recall in the project hypotheses as measures of prediction performance. This makes evaluation part of the proposed work rather than an undefined decision at delivery. section 3.5
During testing, decide which measures matter for the use case. Consider precision alongside recall, F1 score, other performance measures, and nonfunctional results; accuracy alone gives a poor view of performance. Record the basis for the eventual model-selection decision, including trade-offs that testing does not resolve. section 8.6
