Description
Feeds each respondent's conjoint utilities into competitive what-if scenarios to estimate shares of preference, demand curves, price premiums and cannibalisation.
The simulator turns conjoint part-worths into simulated choices, like having every respondent in one room voting on products in a competitive scenario. Each product is defined by the study's attribute levels; utilities are summed per respondent and a choice rule converts them into shares: first choice (highest utility wins), logit or Bradley-Terry-Luce share of preference, or randomised first choice. Shares are averaged across respondents with confidence intervals. From a base case the analyst runs one-attribute-at-a-time sensitivity analysis, demand curves and elasticities, feature value (the price rise that returns an enhanced product to its base share), line extensions by segment, and optimisation. Louviere adds calibration to observed choices and the inclusion of a no-purchase option or current state to forecast demand. Orme stresses that it predicts preference, not market share.
Attributes
- Discipline
- Product Analytics & Strategy
- Loss type
- Siloed
- Stage
- Strategy, Definition
- Data
- Quantitative
- Effort
- Weeks
Steps
- Obtain individual-level part-worths from a conjoint study (for example CBC with hierarchical Bayes).
- Define a base-case scenario of current competitors using only attribute levels measured in the study, plus a no-purchase option where possible.
- For each respondent, sum part-worths per product and apply a choice rule: first choice, logit share of preference, or randomised first choice.
- Average across respondents to get shares of preference with 95% confidence intervals (share plus or minus 1.96 standard errors).
- Run sensitivity analysis by changing one attribute of the target product at a time with competitors held constant.
- Build demand curves by varying price across measured levels, and compute elasticity by the midpoint formula or log-log regression.
- Value features by adding a feature and raising price until share returns to base; simulate line extensions by segment to weigh gain against cannibalisation.
- Optionally calibrate to observed recent choices with a multinomial logit, without tuning to market share in ways that distort elasticities.
Sources
- Analyzing Decision Making: Metric Conjoint Analysis (Louviere, 1988)
- Getting Started with Conjoint Analysis (Orme, 2010)
- 9 Things Clients Get Wrong about Conjoint Analysis (Chapman, 2013)
- Uncommon Choices: Novel Applications of Conjoint Analysis in Practice (Chapman, 2021)
- Conjoint Analysis: Predicting What Users Really Want (Jozranjbar, n.d.)
- R for Marketing Research and Analytics (Chapman & Feit, 2019)
Also known as
- Market simulator
- What-if simulator
- Share-of-preference simulation
- First-choice simulation
- Conjoint market simulation
- Multi-store simulator
Categories
- Analytics & Behaviour
- Product Analytics & Strategy