Product extras such as optional services, extensions or enhancements are a convenient way to increase the overall amount the customer is willing to pay. The cost increase, if not recognized by customers as bringing an expected value, might have an opposite effect and decrease the demand. Thecorrect price setting of options is, therefore, an objective in the aim to increase the core product attractiveness and the amount customers spend once they start to buy.
DCM-based models are well behaved under the assumption of just one item being chosen from a choice set. In real life, products are often bought concurrently in various quantities on different occasions and for specific reasons. Complementarity of the items, affordability, willingness to spend (budget limits) and saturation effects (needs, expectations, consumable quantity limits) play an important role. The related demand has been successfully described by aggregate metric models on the market scale level. In marketing research, these models are not quite usable as the prompted statements about consumed quantities are often questionable. However, the situation is much more favorable in case of options purchased asn single units. A DCM approach can be applied.
A purchase of some quantity of a product can be modeled as repeated choices with the same probability of each choice. A purchase of some quantities of several products concurrently can be modeled as a sequence of multiple independent choices. This approximative model that combines the initial and repeat choices into a single event has been named "volumetric CBC" by Sawtooth Software, Inc.
If some products are purchased concurrently as single-unit components that make up a single purchase, the model is much simpler as no "repeat choice" needs to be considered. A simulation of such activity is known as MBC - Menu Based Choice. The focus is on asking respondents to configure their preferred choice by making from none to multiple selections from a menu of possible selections.
|Customers are often interested in a narrow selection of options some of which would be missing if only a subset of items were offered. To simulate a realistic situation, all optional items as considered "on the menu" should be shown in all choice sets. The conventional CBC in chip allocation arrangement is appropriate for testing concurrently purchased items. It has been used in a range of mass customization studies, such as|
On the assumption of a fixed expenditure (budget) the behavioral model is formally identical with the indirect utility model derived from the CES - Constant Elasticity Substitution or CDES - Constant Difference Elasticity Substitution utility models. If an upper bound on the total chosen quantity is assumed (e.g. if options are substitutes) the approximate equivalence is achieved when an outer product is formally added into the model. The results can be presented in any form available for a conjoint study (utility, acceptance, elasticity, potential, simulation, etc.).
The arrangement, typical for the MBC - Menu Based Choice method using a standard CBC designer, is not without problems. Due to the pseudo-random construction of level assignments in an orthogonal design, some price levels are decreased and other and other levels are increased. The sum of prices changes substantially less. Respondents recognize this very early, and, at least with some types of items on the menu, show no or only small changes in preferences with the randomized prices of the options except the most expensive options. This behavior is typical for choices governed by a willingness to spend, for a strategy "must have" or "no need for", and for menus where most items do not have possible substitutes. One can appease hunger with a donut instead of a hamburger but not be satisfied with a car having towing equipment instead of electric windows. For the purpose of bundling it may be more appropriate to search for a sensitivity of the total price (computed for the chosen options) rather than to try to estimate the price sensitivity for each of the options independently. This can be achieved by a modified MBC arrangement with parallel rather than randomized changes in item prices.
A modified MBC test starts with all prices at the highest prices and, in each of the consecutive steps, all prices are decreased in parallel by predefined increments, typically a constant percentage. The number of choice tasks is equal to the number of price levels, usually 5 to 7. The actual setting of the levels depends on the assumed econometric model and the expected influence the changes may have on the choice behavior.
The experimental arrangement of a modified MBC test resembles the BPTO - Brand price Trade-Off method, several Gabor-Granger price sensitivity tests done in parallel, or a BYO - Build Your Own user design. It has possibly inherited some of the virtues and vices of the mentioned congeneric methods. The results are supposed to be sufficiently informative as they are obtained in an experimental arrangement similar to a real situation such as an imitation of a "discount on everything". The probabilistic DCM - discrete choice model approach and use of Bayesian type of the model parameters estimation adds to robustness and is believed to reduce estimation bias. CDES - Constant Difference Elasticity of Substitution is the underlying utility model that allows estimation of an aggregate price elasticity for each option provided sufficient data are available. The results are probability-based and can be merged with the results of any other DCM-based method, typically a CBC or SCE of the core product the options are offered with.
|The main result of an MBC test, be it
the standard or a modified one, is the choice probability of each
option given by its properties (typically price) and the person.
There are two basic approaches to the design and analysis of an MBC
In case the items in an MBC test are selectable options or accessories offered as an extension to a core product, the influence of the basic features of the core product should be known or evaluated independently from MBC. It is usually the product and its properties that have the highest influence on the acquisition and/or retention of customers. The number of the basic features and selectable options is often so high that the exhaustive testing of all their combinations separately is intractable. However, the options can be made into several (a priori most influential) packages and included in the test of the core product as levels of an additional attribute. These "representative" packages serve as "group levels" of the available options and their attractiveness. In this way, it is possible to assess the influence of the availability of the options on acceptance of the core product. The selection of the representative options and their presentation in the test is critical and must be done with the utmost caution and expertise. If their design conforms to the statistical rules of DOE (Design of Experiments) the obtained data can be merged with the data from an MBC test aimed at the options alone.
The stated demand for the options obtained from an MBC is assessed given the core product acceptability. The test of options alone will not provide information on a possible change in the attractiveness of the core product. The main advantage of an MBC lies in the ability to test a large number of options. Up to 25 options have been tested without a noticeable decrease in performance. It is probable that the number can be even higher without the excessive deterioration of the reliability of estimated parameters.
The applications we had carried out so far dealt mostly with options of banking and telecommunication products and services and were used as extensions of the conjoint market simulators. They cannot be presented here for apparent reasons.
Combining multiple items into bundles is a common strategy when preferences and willingness to pay (WTP) vary across individuals. In turn, offering multiple items individually may yield the maximum profit. MBC is a natural method of help in the decision of which of the approaches to choose.The options included in each of the packages must address a substantial fraction of the potential target for the core product. Segmentation of the target by the expected interest in options can be done with LCA - Latent Class Analysis. To find the most appealing package (or packages) for each of the segments the portfolio optimization is a possible way.
Compared to the idea of "select from all possible aspects and combine them" used in product configurators, a low number of ready-made packages makes the decision for a customer much easier. Such packages should selected from several properly tested packages. Composition of packages is often restricted as not any aspect can be combined with any other aspect. The MBC test method is not suitable. The dual-MaxDiff-based package test can be used with the advantage of providing a view of the influence the aspects of the packages have.