Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

What is difference between ACBC and MBC?

What is the pro and cons for both method?? And what is the best situation to use ACBC instead of MBC? and What is the situation that should use MBC rather than ACBC?
asked May 20, 2017 by liewfk (145 points)

1 Answer

+1 vote
Adaptive CBC is appropriate for situations involving about 5 attributes or more, where buyers consider multiple pre-designed products or services and select just one.  They cannot customize the product.

Menu-Based Choice is appropriate when buyers face a menu of available elements that could make up a total product or service and they select from one to multiple options (elements or offerings) to build their choice.  They can customize their choice: e.g. include a, b, and c to build their preferred choice.  Sometimes menu-based choice involves choices between pre-defined bundles or items that can be chosen a la carte.

Examples of  MBC include: dinner menus, customizing a technology product such as a laptop computer on a website for purchase, single or multiple drug therapies, singe/double/or triple play data services into a home.

To summarize: ACBC is a single choice among multiple product alternatives (single choice using radio buttons).  Menu-Based Choice involves potentially multi-choice (multiple selections) across check boxes.
answered May 20, 2017 by Bryan Orme Platinum Sawtooth Software, Inc. (191,140 points)
Thanks Bryan,

If I want to determine the importance of each attribute and the desirability of each level (options from attribute), I should use ACBC or MBC?
Importance of attributes is typically a function of ACBC.  Preference scores (utilities) for levels of each attribute (or item in the case of MBC) can often be done with MBC as well.

MBC is an advanced approach for experienced choice modelers.  It typically involves much more time, expense, and sample size than the more standard conjoint methods like CBC and ACBC.
May I know why it involves much more time, expense and sample size?
MBC needs to be programmed manually using your own HTML skills and typically JavaScript and some CSS if you want it to lay out pretty.  Sawtooth Software's MBC software program does not create the data collection instrument for you.  You need to program it using "Free Format" questions in our Lighthouse Studio software or using another piece of survey software of your choice.

And, because MBC software doesn't program the survey automatically for you, it doesn't automatically prepare the data for you for utility estimation either.  The data needs to be prepared in a .CSV file, but there is typically some data manipulation and cleaning to be done to take the data out of your survey package and prepare it for MBC analysis.

Then, with MBC models there is so much flexibility and power put on the user to create the right models.  We provide guidance of course in our manual.  But, it's much like having a tool that can run regression analysis: how to formulate your models, code your variables, etc. is left up to the user.  MBC handles much of the details, but the responsibility is on the researcher to specify the right models that fit the data well.

MBC models can be quite demanding statistically on sample size because you often want to incorporate the idea of multiple items being chosen synergistically.  And, you often want to incorporate some cross-effects in the model.  Sample sizes of 1000 to 3000 are not uncommon so that the models stabilize and predictions are solid.

In general, carrying out an MBC study is going to take 5x to 10x more analyst time than a straightforward CBC study.  And, the decisions to be made are often more complex than standard CBC.
Again,  thanks for your detail explantion. In term of output, besides utility score and simulator,  what else mbc can provide to us?  Any difference between acbc and mbc in term of the output that we can get?
As you realize, both techniques lead to a market simulator: a what-it tool that lets you specify competitive product scenarios (or menu scenarios in the case of menu-based conjoint) and predict the likelihood of choice for each of the competitive alternatives in the scenario.

I don't think the focus should be on what MBC analysis can give you versus CBC analysis (as the outputs resemble one another...though with more moving pieces and often more complicated under MBC).  I think the main consideration is whether a menu-based choice (multi-select) approach better mimics the real world purchase decision than a discrete choice (CBC) task.  And, whether the budget and available sample size can support the MBC approach.
Bryan,  thanks for your time.  I think i get the idea.