(for consulting in ML) The Second Rule of Machine Learning - Start Machine Learning with simple shallow models. 50% of the problems are solved with good data

(for consulting in ML) The Second Rule of Machine Learning - Start Machine Learning with simple shallow models. 50% of the problems are solved with good data choice of data + some generalized linear model. 30% remaining problems solved with shallow models or old school ML models. Anything from support-vector machines, decision trees, nearest neighbors, very shallow neural networks. Remaining 20% require more work.