What is ASM (Advanced Simulation Module)?
In typical market simulations, marketers specify a given test product and a competitive context and ask "How good would this product be?" With optimization problems, what they really want to know is "What product would be best?" The Advanced Simulation Module (ASM) extends the capabilities of our popular conjoint market simulator within the SMRT Platform to offer such product optimization.
If product optimization is the goal, you can save a great deal of time and find potentially better solutions using automated searches rather than the manual process of modifying products and processing simulations one at a time.
The ASM can analyze conjoint data from any of Sawtooth Software’s conjoint systems (ACA, CBC, or CVA), including full-profile, partial-profile, and alternative-specific designs. You can even use conjoint or preference data provided by the researcher that was not necessarily generated by Sawtooth Software’s systems.
Conjoint simulators provide an exceptional tool for product optimization. They can take into account the characteristics of currently-available products as well as the desires of a heterogeneous population of potential buyers. Subject to reasonable caveats about the quality of respondent sampling and questionnaire design, conjoint simulations can accurately assess likely product success long before a product is ready for test market. The ASM can optimize based on maximum utility, purchase likelihood, market share, revenue or profitability. Profitability optimization requires additional user-provided information about feature costs. Also, if you include feature costs, the ASM can perform cost minimization searches, searching for products that meet some threshold of utility, share, revenue, or profit while minimizing cost.
The ASM uses heuristic search strategies to find the optimum (or near-optimum) product or a portfolio of products. It does this by exploring the "response surface" of the objective, such as share, corresponding to attribute levels for the product(s) of interest. Because different optimization strategies are more effective with different kinds of response surfaces, the ASM provides several from which to choose.
Five optimization algorithms are available:
Exhaustive Search is the simplest of the algorithms. It examines all possible combinations of the product or products to be optimized. The main strength of Exhaustive Search is that it is guaranteed to find the best solution. If the response surface has multiple peaks, Exhaustive Search will evaluate all of them and choose the best one.
The main shortcoming of Exhaustive Search is that the number of combinations to be evaluated can become very large. For most product optimization problems, Exhaustive Search will probably not be feasible until the search space has initially been reduced using other methods.
Grid/Gradient/Stochastic Search are three "hill-climbing" methods that are much faster than Exhaustive Search. The only thing different about them is the way they deal with interpolable (continuous) variables. These hill-climbing methods can find near-optimal solutions typically within a few hundred simulation steps. If using a fast simulation method (such as Share of Preference or First Choice), one typically obtains a solution within about a minute. If using a more intensive simulation method such as Randomized First Choice, it usually takes a few minutes to solve.
The main strength of these hill-climbing methods is their speed. If the response surface is single-peaked, hill-climbing procedures are guaranteed to find the optimum. If there are multiple peaks, then repeated runs from different starting points are quite likely to find it.
Genetic Search uses ideas from evolutionary biology. Each product is considered a "chromosome" and each attribute a "gene." We generate of pool of possible solutions obtained randomly. In each iteration ("generation") the least "fit" half of the population is replaced with new members obtained by random "mating" of the most fit members. Each child has a combination of the parents’ characteristics (genes). Also, the child’s attribute levels are subjected further random "mutation." One strength of Genetic Search is that it makes no assumptions about the shape of the response surface. It is theoretically possible that the final population may include near-optimal members who are high on different peaks, and in these cases Genetic Search is less vulnerable to local optima. Genetic Search’s main shortcoming is slowness. We have found it to take five to ten times longer than the faster methods.
The Advanced Simulation Module can be used in a variety of contexts to solve a variety of optimization problems:
- New product introduction without competition
- New product introduction with competition
- Optimizing multiple products simultaneously
- Line extension optimization
- TURF-like problems (find minimal set of products to maximize appeal)
- Maximize appeal subject to a maximum cost
- Minimize cost subject to meeting a performance threshold
These examples, and the details surrounding this new module, are described more fully in the Advanced Simulation Module Technical Paper.