An independent study reference written by Dr Phuc V. Nguyen. It is not official subject material — for assessment requirements always follow your subject outline and vUWS.
The decision process
Herbert Simon described deciding as three phases. Intelligence is noticing that a problem or opportunity exists and gathering what is known about it. Design is generating alternatives and working out what each would involve. Choice is selecting one and being able to defend it. Later writers add implementation and review, since a decision never carried out or never checked is unfinished. The phases are not a queue. Work loops back constantly, and the loop from choice into design is where most of the value appears. Analytics attaches differently to each phase, and attaching it to the wrong one wastes projects.
Why it matters
Think of a doctor. First they work out what is actually wrong, which is not the same as listening to the complaint. Then they lay out the treatments that could plausibly work. Only then do they pick one, and afterwards they check whether the patient improved. Skip the middle step and you end up treating the first idea anyone said out loud rather than the best one available.
A team is handed the request "predict which customers will leave next month". Before any modelling, the most valuable question is
Formulas
Worked examples
A supermarket group asks for a forecast of next quarter demand across its bakery range. The team delivers an accurate forecast. Nothing changes.
The request arrived phrased as a choice problem when the real gaps were in intelligence and design. Nobody had established what the bakery problem was, whether waste, stockouts or margin, and nobody had listed the actions available, which are baking schedules, order cut-off times and markdown timing. A forecast attaches to none of them. Rerun from the start and the useful output changes shape entirely: waste by store and by hour, mapped to the two or three schedule changes a store manager can actually make.
An operations team has spent four days debating the two supplier options that were on the table at the first meeting.
The debate is stuck in choice because design was skipped. Two options that both look mediocre usually signal an option set that is too small. An hour spent generating alternatives, splitting the contract, renegotiating with the incumbent, staging the switch over two quarters, is worth more than a fourth day of comparing the same pair. Analytics helps here by costing new alternatives quickly, not by refining the comparison of old ones to another decimal place.
Common mistakes
- ✗The process runs once, in order, from start to finish. It loops. Evidence found during choice routinely sends you back to design, and a badly framed problem sends you back to intelligence. Looping is how the process works rather than a sign that it failed.
- ✗Analytics belongs in the choice phase. Most of its value sits earlier, in framing the problem and in generating and costing alternatives. A model that only compares options someone else wrote down inherits every gap in that list.
- ✗A decision ends when the choice is announced. Implementation and review are part of the process. Without review there is no feedback, and a team that never checks its past calls has no way to improve its future ones.
- ✗Good decision makers consider every option. Simon's point is that nobody can. Search costs time, so people stop at the first alternative that clears an acceptable standard, which makes the order in which options surface surprisingly powerful.
Revision bullets
- •Simon: intelligence, design, choice; later work adds implementation and review
- •Intelligence asks whether there is a problem and what is already known
- •Design builds the option set, and is where analytics is most underused
- •Choice selects and justifies, then the decision is carried out and checked
- •The phases loop, and going backwards is normal
- •Satisficing: stop at the first option that clears the aspiration level
Quick check
A team is handed the request "predict which customers will leave next month". Before any modelling, the most valuable question is
Two supplier options have been argued over for days and both still look mediocre. The best reading of that stalemate is
Connected topics
More in Decisions and Models
Sources
- Simon (1977)Simon, H. A. The New Science of Management Decision. Revised edition, Prentice-Hall, 1977.Sets out the intelligence, design and choice phases, and argues that design receives the least attention of the three in practice.
- Simon, H. A. "A Behavioral Model of Rational Choice." Quarterly Journal of Economics, 69(1), 1955.Introduces satisficing: choosing the first alternative that meets an aspiration level, rather than searching for the best.
- Sharda, Delen & Turban (2020)Sharda, R., Delen, D., & Turban, E. Analytics, Data Science, and Artificial Intelligence: Systems for Decision Support. 11th ed., Pearson, 2020.Standard treatment that extends Simon's three phases with implementation and monitoring, and maps analytics support onto each phase.