Analytica

Every day I find it more useful and useful for all aspects involving data processing and analysis for decision making.

I have been an avid and consistent user of Analytica since it became publicly available in 1996 (actually before that when it was called Demos), and it's not inaccurate of me to say that I use it almost daily as part of my analytic activities and services.

There were two main features that originally attracted me to Analytica: its influence diagram interface for graphically laying out the flow of model logic, and its Intelligent Arrays technology which permits array abstraction. Both of these features integrate with each other to give a modeler the ability to structure and communicate model logic in the most concise and informative way possible. The added benefit of the Intelligent Arrays is that it supports the extension of model dimensionality with little recoding, if any at all. The other indispensable features of Analytica are its support for Monte Carlo simulation, built-in graphing tools, and support for connecting to external data sources. By comparison to a spreadsheet, models of similar complexity require anywhere from 1/4 to 1/2 the amount of time to build in Analytica.

Analytica really excels with communicating complex ideas. The graphical layout of the influence diagrams enhanced with control UI elements supports clients' more readily comprehending important model details, especially if they were not directly involved with the creation of the model. I frequently take executives through very detailed discussions of risk analysis without preparing post processed and simplified information or additional slide decks.

The one drawback that modelers can run into is that the Intelligent Array feature can lead to significant memory and computational time requirements in particularly dimensionally complex models or models that use a significant number of Monte Carlo trials. However, Lumina provides some profiling tools that allow a modeler to find the sources of processor and memory resource burdens; therefore, in many cases a modeler can find ways to improve the efficiency of their code. The effect of requiring significant amounts of memory or run time of models is not something that occurs often. The recommended memory and CPU capabilities'readily available on most business computers'will exceed most users' requirements.

The combination of being a powerful analytical tool, and at the same time having a high capacity of visual or diagramal transmission of information allows us to use this program to make decisions, transmit and convince more easily who will ultimately make the final decision. which often leads to higher rates of return on investment that will be made.

Analytica encourages good design and comprehensibility . It can generate programs structured much like the underlying mathematics. More concretely: (1) its visual modeling is powerful for design, communication, and maintenance; (2) its use of arrays makes it possible to deal with multiple objectives, distinctions, and uncertainties elegantly and compactly; (3) it has many built-in features for probability and statistics, and for Monte Carlo analysis; and (4) all of this is great for uncertainty analysis.

Analytica is not well known, limiting portability. Most programmers will not find it intuitive at first because it is so different. Analytica is not well suited for connecting models or modules built in different languages. It is not suitable for agent-based modeling or for discrete-event simulation. Analytica's graphics are good for basic purposes, but not as versatile as they might be.

It is possible to learn a great deal about and to actually use Analytica with a few days or a week of serious effort. Using the tutorial is essential.

Advanced features are more challenging (as in all languages). Fortunately, Analytica allows "procedural programming." For many/most models, such programming undercuts simplicity and clarity, but for more complicated models, it may be more practical.