🛑STOP assuming that the quality of business decisions depends on data quality and the accuracy of algorithms #ai #ml #LLMs.
💢The problem with many #data and #AI leaders is that they think primarily about the “potential” impact of the quality of their work on business decisions. This leads to overrating the quality of data and the accuracy of the algorithms they build. They safely assume that if they improve the quality of data or accuracy of the algorithms, consequently, they enhance the quality of business decisions, which is a wrong assumption as they are not based on a concrete decision model.
💣Business decisions are complex and happen hundreds of thousands of times daily. To improve their quality, we need to know exactly how they are made in the first place. The outcomes of business decisions are more than likely the result of multistep decisions. That means the outcome of one decision is the input of another decision.
To understand how business decisions are made, we have to frame them using decomposition techniques. Let me explain it in principle using Decision-Centric Approach® here:
🔹Peek a business operation (function, group, team, value chain…)
🔹Understand and identify what “business decisions” they make that influence the operation
🔹Decompose a business decision into smaller, easier-to-understand decision units (e.g. other smaller business decisions)
🔹Establish the dependencies between those decision units
🔹Identify and determine the Metrics and KPI for each of the business decisions (e.g. decision units)
🔹Identify the data requirements for each of those decision units
✨At this point, you have depicted a multistep and hierarchical decision model (i.e., a Decision Graph) that represents a holistic view of business decisions. Now, you can use this model as a blueprint and roadmap of business decisions and determine what improves the quality of them as a whole. #decisioncentric
🚀The quality of business decisions depends heavily on the outcome of each decision unit, as part of the decision graph above. Improving the quality of each of those units requires different techniques, as by nature, not all of them are based on data, and they are not all machine learning models. More likely, they are a mix of everything, including, but not limited to, #businessrules #optimization #dataprocessing #machinelearning, and so on.
#decisionintelligence
Posted here.
Published May 13th, 2024 at 07:30 am
