When building analytics solutions, maximize the quality of your data and minimize the complexity of your algorithm.
The data should be made up of only the most influential factors. Each input source needs to robust and reliable. This means investing in solutions for monitoring and dealing with missing values and outliers.
The algorithm should be the simplest approach that gives you good accuracy. Trading a few percent of accuracy for simplicity is usually a good option.
This pattern enables your domain experts to build trust in the solution, supporting the ability to reason about the recommendation provided by the solution. This leads to optimal data driven decision making, knowing when to trust the system and when to intervene.