Image source: marchoeijmans.blogspot.com
Operational research (OR) analysis is all about modeling operational situations and problems with math, logic and scientific theory – or is it? Despite the common perception of scientists in lab coats (operational research would like to get them into business suits, but that’s another story), heuristics (‘doing it by feeling’) can also be a viable approach. Because it’s often easier to modify a concept that exists rather than invent something from scratch, the rules of thumb that characterize heuristics can provide a valuable shortcut to a first approach, saving time and energy. But at which point do you then need to swap heuristics out and rigorous OR analysis in?
Bounded rationality and satisficing
Heuristics may get you started, but they are unlikely to lead you to an optimal decision. Left to their own devices, people may well lean towards ‘satisficing’, the term invented by Herbert Simon in his studies on problem solving. Satisficing describes the acceptance by people of solutions or situations that they find are good enough for them, without worrying about optimizing them further. This ‘bounded rationality’ (another term invented by Simon) can be found in business contexts as well: a software release that is ‘sufficiently bug-free’, a contract that is ‘something we can live with’, a business negotiation that ‘doesn’t leave too much money on the table’, and so on.
Computer programs that use heuristics
Humans are not alone in using rules of thumb. Computer virus programs do it as well (even if the heuristics they are programmed with originally came from human brains). The programs compare suspect code with past experience to detect possible matches. As the library of past experience grows, the program becomes more adept at spotting virus danger. However, it often remains incapable of detecting new viruses that operate in other ways. Likewise any other software that relies exclusively on past experience to guide conclusions about the future will be limited in its capabilities and may completely miss new trends and developments.
Defense against blinkered thinking
If the use of heuristics means a risk of OR analysis myopia, tunnel vision or similar phenomena, what is the solution? A popular technique to expand understanding and awareness is brainstorming, in which participants give unbridled expression to any ideas they may have about solving a problem, however far-fetched those ideas are. This can help to break through any self-imposed limits due to an approach based on heuristics, although it is still no guarantee of a complete or optimal solution.
You’re wrong anyway, but how about being less wrong?
Satisficing in OR analysis may come about because of a deliberate decision to stop the analysis once a certain level of results has been achieved. Some projects demand this: software development cannot realistically be tested to 100 per cent in terms of branch coverage (different routes through the code), so another figure must be chosen for a sufficiently good level of test. In a sense, satisficing is built into the very idea of modeling in that ‘all models are wrong, but some models are useful’. Models become more useful as they are modified to converge on more credible, realistic behavior and results. To be more wrong with less effort, it makes sense to choose a modeling solution that allows for easy model adjustment such as Analytica with its Intelligent Array technology.
If you’d like to know how Analytica, the modeling software from Lumina, can help you with OR analysis, then try a free evaluation of Analytica to see what it can do for you.
