--- title: "Run Conjoint Analysis with AI | Minds" canonical_url: "https://getminds.ai/use-cases/conjoint-analysis-with-ai" last_updated: "2026-08-25T03:49:23.534Z" meta: description: "Configure attributes, run server-built D-optimal choice designs, and simulate product preference shares using the conjoint analysis workflow in Minds." "og:description": "Configure attributes, run server-built D-optimal choice designs, and simulate product preference shares using the conjoint analysis workflow in Minds." "og:title": "Run Conjoint Analysis with AI | Minds" "twitter:description": "Configure attributes, run server-built D-optimal choice designs, and simulate product preference shares using the conjoint analysis workflow in Minds." "twitter:title": "Run Conjoint Analysis with AI | Minds" --- Minds July 31, 2026·Use-case·Minds Team # **Run Conjoint Analysis with AI** Minds runs executable conjoint analysis to estimate attribute trade-offs and simulate preference shares across product configurations. Use directional synthetic evidence to refine study design before fielding. [Run a Conjoint analysis study](https://getminds.ai/?register=true) Minds provides a conjoint analysis workflow that evaluates multi-attribute product trade-offs using directional synthetic panels. Teams input attributes, levels, constraints, and task parameters to generate server-built D-optimal choice designs. Minds collects choices between complete configurations, fits conditional logit models, and outputs attribute importance, per-level part-worths, holdout validation diagnostics, and directional preference share simulations to guide product strategy before human fielding. ## The decision Conjoint analysis supports Product and research leadership face difficult trade-off decisions when designing software packages, consumer goods, hardware tiers, or service bundles. When a product team needs to determine whether target buyers value extended battery life over a thinner chassis, or lower base prices over included premium support, direct rating scales often fail. Respondents routinely mark every feature as critical and every price point as too high. Conjoint analysis resolves this by collecting choices between complete configurations composed of varied attribute levels. The method quantifies how much value buyers place on individual attributes and calculates the relative utility of each level. By observing choices across structured task sets, research teams estimate per-level part-worths that reveal what respondents sacrifice to secure their preferred features. Conjoint analysis supports portfolio decisions, feature inclusion lists, pricing tier boundaries, and competitive positioning. Adjacent methods like MaxDiff measure the relative importance of individual features in isolation, while Gabor-Granger and Van Westendorp focus strictly on single-item price elasticity. Conjoint analysis evaluates features and prices simultaneously in unified product configurations. Rather than asking respondents what they want, the method evaluates choices between complete configurations across realistic product profiles. In Minds, conjoint analysis serves as an early stage direction engine. Product managers test attribute combinations against synthetic audiences to identify feature bundles and pricing structures before launching human fieldwork or engineering initiatives. ## Configure the study Executing a conjoint study in Minds requires defining structural design parameters before publishing the run to an audience. The system relies on specific inputs to build the choice experiment. First, specify the attributes and levels. Attributes represent the core dimensions of the product, such as screen size, monthly price, battery runtime, or warranty length. Levels are the discrete values each attribute can take. Second, define feasible constraints. If certain attribute level combinations are logically impossible or commercially non-viable, such as a basic tier featuring enterprise security controls or a high-performance model set to the lowest price point, input feasible constraints. Minds excludes prohibited combinations from the design space. Third, set the task count. Specify a task count between 4 and 40 to determine how many choice sets are evaluated during the run. Fourth, define alternatives per task. Specify between 2 and 5 complete product configurations presented in each choice set. Fifth, configure optional holdouts. Set between 0 and 10 optional holdout tasks. Holdout tasks are choice sets set aside during configuration that are excluded from conditional logit model training. Minds uses holdouts solely to produce validation diagnostics after fitting model parameters. Sixth, select the estimator id, which is configured as multinomial-logit-v1. Once attributes, levels, feasible constraints, task count, alternatives per task, optional holdouts, and the estimator id are submitted, Minds constructs the study foundation. ## How Minds runs the method Execution follows a structured pipeline. Once parameters are finalized, the server builds a D-optimal choice design. This procedure constructs an experimental design that respects all defined feasible constraints and task limits. The resulting matrix balances level presentation and optimizes statistical estimates. Minds then deploys the choice design to the defined synthetic audience panel. Synthetic profiles execute choices between complete configurations across the generated task sets. In each task, the synthetic respondent evaluates the side-by-side product concepts and selects a single preferred option. Forced choices mirror the trade-offs buyers confront in procurement environments. After collection completes, Minds initiates part-worth utility estimation using the conditional logit estimator. The conditional logit model processes choices between complete configurations to quantify the relative contribution of each attribute level to overall profile preference. Per-level part-worths are computed across all attributes, converting categorical selections into precise continuous log-odds metrics. The pipeline next runs the holdout validation stage. The registered holdout validation stage produces diagnostics by testing the model on the holdout tasks set aside during configuration. Minds records hit rates and diagnostics to establish baseline fit. Finally, the platform executes preference share simulation only when configurations are supplied. The system models competitive scenarios by presenting target product configurations alongside alternative concepts, using the estimated part-worth utilities to project directional preference shares. These outputs return directional preference shares and are never called forecast market share. ## Interpret the output The output generated by a Minds conjoint run delivers structured numerical analytics tailored for research synthesis. The choice design table documents the complete matrix of tasks presented across the study, confirming design balance and constraint compliance. The raw choices dataset contains selections recorded across all presented task sets, providing an audit trail for downstream verification. The per-level part-worth estimates summarize the utility of every attribute level, organized by separated level ids. Positive part-worth values indicate increased preference relative to the baseline level, while negative values reflect relative disutility. Research teams evaluate these numbers to determine the calculated preference drop when shifting between pricing tiers or feature inclusions. Attribute importance scores express the relative contribution of each attribute to total choice variation. This identifies which dimensions drive choices overall. An attribute importance of forty percent for price indicates that pricing exerts twice the influence on decision outcomes compared to a warranty attribute scoring twenty percent. Fit output parameters contain log likelihood, null log likelihood, McFadden R-squared, hit rate, observations, parameters, convergence status, and separated level ids. Validation diagnostics from the holdout validation stage deliver diagnostic performance metrics, including hit rate on holdout tasks. When configurations are supplied, simulation outputs return directional preference shares. These show how directional preference shares shift when changing feature configurations or price points. Teams test hypothetical product launches against competitor baselines to identify configurations that maximize directional preference shares. These directional preference shares are never referenced as forecast market share. ## Workflow for professional market-research and product teams Professional research and product organizations integrate Minds conjoint runs into discovery and validation pipelines. Teams begin by defining candidate attributes and prospective pricing tiers with internal stakeholders. Research managers enter these parameters into Minds to run synthetic pre-tests. Synthetic panels provide directional feedback on trade-off sensitivity, allowing teams to iterate. If initial part-worth outputs show that certain features contribute negligible utility, product managers eliminate those attributes or adjust level boundaries. If price sensitivity dominates all other dimensions, teams recalibrate level spacing to establish clearer differentiation. Once the synthetic conjoint run yields clear attribute importance, separated level ids, and holdout validation diagnostics, teams export the finalized experimental parameters. The attributes, levels, and constraint rules are then deployed into recruited human panel surveys. By pre-testing the conjoint design in Minds, researchers evaluate task parameters, review fit metrics like McFadden R-squared and log likelihood, and refine survey structures before human fieldwork. The directional synthetic output informs executive briefings and aligns product roadmaps, while human recruitment delivers external validation. For post-launch management, teams re-run simulations in Minds by supplying new configurations to evaluate directional preference shares without re-fielding live studies. If a competitor drops prices or introduces new bundles, analysts adjust profile parameters in Minds to simulate directional preference shares before committing resources to a formal field study. ## Limits and validation Synthetic audience evidence in Minds is inherently directional. Conjoint runs executed on synthetic panels reflect modeled persona behavior based on underlying baseline groundings, not statistical market truth. Minds does not claim statistical representativeness or universal accuracy. Synthetic conjoint studies do not replace recruited human respondents, real-world panel verification, or true behavioral purchase data. Instead, Minds complements real fieldwork by screening hypotheses, identifying attribute ranges, improving questionnaire structures, and focusing recruitment budgets on research questions. Certain commercial edge cases require validation with recruited human respondents. High-stakes pricing changes, novel product category launches without historical market precedent, and regulatory reviews must be verified through real human panels. Synthetic panels provide directional screening, but human fieldwork remains required for final risk mitigation. By executing conjoint analysis in Minds as an iterative design tool, research and product teams reduce survey development timelines, evaluate model parameters, and enter human fieldwork with structured instruments. ## **Frequently asked questions** ### **Can Minds run conjoint analysis end to end?** Yes, Minds executes registered conjoint analysis within the platform. The system generates D-optimal choice designs, collects forced-choice responses, fits conditional logit models, and outputs share simulations. All calculated outputs remain grounded in directional synthetic audience panels. ### **When should a team use conjoint analysis?** Use conjoint analysis when you need to evaluate trade-offs across multi-attribute product concepts and predict feature selection. It is ideal for optimizing packaging, subscription tiers, and feature bundling before committing engineering or survey budget. ### **What inputs are required to launch a conjoint study?** You must supply structured attributes and levels, feasible constraints, task count per respondent, number of alternatives per task, and optional holdout tasks. Minds uses these parameters to assemble the experimental design automatically. ### **Does synthetic conjoint analysis replace recruited human research?** No, synthetic audience results are directional and do not replace recruited human respondents. Teams use Minds to screen configuration concepts, eliminate weak feature combinations, and focus external fieldwork on high-stakes trade-offs.