After ACBC v8.3, Do We Need CBC Anymore?

Last Updated: 03 Oct 2014Hits: 6906
With the release of SSI Web 8.3 and the updates to the Adaptive Choice-Based Conjoint module to exclude tasks like the BYO, Must-Haves, and Unacceptables, do we even need the CBC module?

Over the past few years research studies both within and outside of Sawtooth Software have generally found ACBC to provide more precise, in-depth understanding of individual-level preferences than CBC. ACBC utilities generally have greater predictive validity than CBC for complex and hard-to predict CBC-formatted holdout tasks, even though methods bias would seem to favor CBC! ACBC is a longer survey, but more engaging and less monotonous to respondents (and clients alike) than CBC.

Now that ACBC 8.3 can do alternative-specific designs, has the option to skip the BYO (Build-Your-Own configurator) section straight into the CBC-looking round-robin choice tournament, it raises the question if it is now a substitute and replacement for CBC.

Despite the prodigious benefits of ACBC—and there are many—here are some reasons you should keep CBC prominently in the toolkit:

  1. Non-adaptive CBC is the global standard worldwide. Many academics and clients expect the tried-and-true methodology. So, a researcher should have both CBC and ACBC ready to go, depending on the client’s background and preconceptions.
  2. Adaptive CBC does tournament-style CBC tasks, where winning concepts move on and losing concepts are eliminated. Many people will have historic interest in and preference for non-adaptive, purely orthogonal designs.
  3. Adaptive CBC cannot do shelf-display CBC so common in FMCG studies. ACBC can display at most 3 concepts per task in the CBC-looking choice tournament section.
  4. Sawtooth Software’s CBC system supports chip-allocation, best-worst choice of concepts within a task, and dual-response none answer strategies. ACBC only offers discrete choice (pick best within each task). The optional Calibration Concept section in ACBC is similar in many respects to CBC’s dual-response none, but not exactly the same. This may or may not matter, unless the client really wants dual-response None for each choice task.
  5. Adaptive CBC cannot do partial-profile CBC studies where all attributes and levels rotate through the choice tasks for every respondent. Adaptive CBC, optionally, has a form of partial-profile that you may implement where each respondent can screen upfront for important attributes and only the important attributes and relevant levels are taken into the core ACBC survey.
  6. If your study involves showing fixed alternatives (such as pharma studies where some drugs are fixed and others have varying attributes), then it may be very important to you (for context-specific utility estimation within the realistic market scenario) that every choice scenario has all (say) 5 drugs available to the respondent. ACBC cannot do this. Once a constant alternative is rejected in the choice tournament (where only 3 concepts at max can be shown at a time), it will never appear again to the respondent.
  7. ACBC typically requires from 1.5x to 3x more time for respondents to complete than CBC (though respondents find it more engaging and less monotonous than CBC to spend that extra time). The demands of the research project may not permit the longer questionnaire. If aggregate analysis is sufficient for you, CBC doesn’t have to place such demands on the respondent’s time. CBC can be done with each respondent just completing one or a few choice tasks (assuming you can afford much larger than normal sample sizes).
  8. CBC software may be used for “paper-and-pencil” studies whereas ACBC (because it is interactive and adaptive) may not. Sometimes, our users set our CBC software into “paper-and-pencil” mode because they are required for a given client (internal or external) to use a non-Sawtooth Software interviewing platform. So, they manually program a few versions (blocks) of the paper-and-pencil questionnaire within the non-Sawtooth Software survey platform. The responses may easily be imported into our CBC software for subsequent data analysis—and CBC doesn’t care whether the data came from paper-and-pencil respondents or from a third-party web-based survey.

I might be forgetting other reasons that ACBC is not a substitute for CBC. Let me know! Both approaches have different strengths, though with the improvements made to ACBC within SSI Web v8.3, no doubt the use of ACBC will increase.

For more information on ACBC, please read the ACBC Technical Paper (2014)