Adaptive Choice-Based Conjoint (ACBC)

Description

A conjoint survey that adapts to each respondent: they configure their ideal product, rule out unacceptable options, then make choices among tailored alternatives.

Standard choice-based conjoint shows everyone the same kind of randomised product profiles, many of which a given person would never consider. ACBC runs in three stages. In a build-your-own section the respondent picks their preferred level of each attribute. A screening section then shows near-variations of that product and asks whether each is a possibility, which reveals must-have and unacceptable levels. Finally a choice tournament asks them to pick between the products that survived screening. Because every task is close to what the respondent actually wants, the data is more engaging and gives better predictions with fewer respondents, and it captures non-compensatory rules ('anything without offline mode is out') that ordinary conjoint models struggle with. The cost is a longer survey and specialist software for design and estimation.

Attributes

Loss type
Siloed
Effort
Weeks

Steps

  1. Define attributes and levels from qualitative research, keeping them concrete and mutually exclusive.
  2. Set up the build-your-own, screening and choice tournament sections in ACBC-capable software.
  3. Pilot with a few respondents to check length (often 15 to 20 minutes) and comprehension.
  4. Field to the target sample, typically 150 to 300 per segment.
  5. Estimate part-worth utilities, usually with hierarchical Bayes, and identify must-haves and unacceptables.
  6. Run market or tier simulations to compare candidate products, bundles or prices.

Sources

  • Feature Prioritization Methods (slide deck)
  • Orme, B., Getting Started with Conjoint Analysis (2014)

Also known as

  • ACBC
  • Adaptive CBC

Categories

  • Analytics & Behaviour
  • Survey & Psychometrics