SIMALTO

SIMALTO, developed by John Green in 1977, is the acronym for SImultaneous Multi Attribute Level Trade Off, an expert system that mimics optical character recognition algorithms, and is used to help solve the Return-On-Investment problem facing manufacturers and suppliers as to how they could make their product or service maximize the increase in customer loyalty and perceived value for the least investment.

SIMALTO does not use OLS (Ordinary Least Squares) or logistic regression to determine the worth of each option level; it is not a linear model. Instead, it parallels the philosophy of “Neural Nets” with an expert system approach. It compares the multitude of respondent information points with known results in specific cases. It incorporates expert system logic, e.g., first priority improvements are more influential on choice than second priority ones.

With SIMALTO an expert system is created for each survey. The model fits the data using an iterative search method in which all the information gathered is given a weight, which contributes to choice. This system uses real life logic such as: · Is the feature option one that the respondent indicated was unacceptable or not? · Is the feature option one that the respondent indicated was important? · Is the feature option higher, lower or the same as the re-prioritized feature option? · Is the total price of the concept higher, lower or equal to the budget, and is it “exceeded” by a pre-determined amount?

It determines what Options are Unacceptable, what Features are Most Important, how to improve the current specification (True Trade-off), and where additional money would be spent.

A typical SIMALTO study would provide data on 1) current perceived performance, 2) unacceptable levels, which if received, might cause a search for alternative products, 3) expected performance, 4) priorities for improving status quo, 5) willingness to pay for improvements, so demand can be set against investment, 6) savings from current performance which would cause least dissatisfaction, and 7) the most important attributes.

SIMALTO is different from Other Choice Approaches in that it examines the whole product / service. Respondents think rather than rate numerous similar scenarios; price is a constraint, not an attribute or catch-all variable; it has built-in checks such as the structured interview task that builds on the previous one(s) and verifies earlier responses; it is not affected by statistical assumptions of regression (conjoint).

Some of the benefits of SIMALTO are: 1. Respondents analyzed as individuals 2. Questioning context is known 3. Features and their options are not evaluated separately 4. Price is included in the preference process at both the options and total product level 5. Collection of information is interesting for the respondent 6. There is a realistic incorporation of various aspects of importance 7. Maximum accuracy in the vicinity of current marketplace offerings 8. No single equation assumed to replicate “customer preference” model 9. General “rules” of the expert system are more likely to be robust than parameters of regression equations with their assumption reliance on feature distribution and correlation 10. Brand Image, Brand Worth and Feature Value are dealt with separately so the analyst can check the expert system forecasts at this level prior to their combination for full product forecasting. 11. Management can refer back to original data, meaningful in its own right, if they are uncomfortable with a forecast.

Green, John “SIMALTO Modelling – Pro and Cons, A personal View”, Research for Today, May 2002