What is decision intelligence?
Decision intelligence is a discipline that treats a recurring business decision as something to design, measure, and improve. It combines data analysis, predictive models, optimization, and business rules with an explicit model of the decision itself: the actions available, the outcomes that matter, the factors that connect the two, and the constraints that apply. It draws on decision theory and management science as much as on machine learning.
The discipline grew out of a common gap. Companies built dashboards and forecasts, yet the decisions those tools were meant to inform stayed inconsistent and were rarely checked against what happened next. Business intelligence reports what happened. Decision intelligence starts from the decision, works backward to the data and models it needs, and checks whether the choice worked.
How decision intelligence works
A team picks a frequent decision with clear stakes, such as how much stock to reorder, which claims to investigate, or what discount to offer a renewing customer. It maps the levers the decision-maker controls, the outside factors they do not, and the results the business cares about. Predictive models estimate how each result responds to each lever, and rules encode policy limits. Prescriptive analytics or optimization then recommends an action. The system logs each decision with its context so the outcome can later be compared with the forecast. Many analytics programs never close this feedback loop, so their recommendations do not improve over time.
Decision intelligence in practice
A regional grocery chain decides each evening how deeply to mark down perishable items in every store. Store managers used to rely on judgment alone. The chain now models the trade-off between waste and margin, forecasts next-day demand by store and item, and applies its policy limits on markdown depth. Managers receive a recommended markdown for each item and can adjust it with a reason. Each week the chain compares actual sell-through with the forecast and retunes the model where the gap is widening.
What to watch for
- Vendors apply the label to products ranging from dashboards to optimization engines, so ask which decision a product improves and how the improvement is measured.
- Start with a frequent, measurable decision, because rare strategic choices give too little feedback to learn from.
- Log the recommendation, the final decision, and the outcome together, or the feedback loop cannot work.
- Agree in advance who owns each decision and when people may depart from the recommendation.