Conjoint Analysis

Purpose

Determine how customers value different product features, price points, and packaging options by presenting trade-off scenarios—forcing respondents to choose between realistic product configurations rather than rating features in isolation, thereby revealing true willingness-to-pay for individual attributes and identifying the most commercially attractive product configuration.

Conjoint analysis addresses a limitation of direct pricing methods such as Van Westendorp (the referenced method): customers evaluate price in context of features. A customer willing to pay 9.99 for a full-featured product may only pay 4.99 for a basic tier. Conjoint reveals these trade-offs quantitatively.

When to Use

Use conjoint analysis when:

• Designing pricing tiers (Basic/Pro/Enterprise) and need to determine which features justify premium pricing.
• Validating feature–price trade-offs: customers value Feature~A at but Feature~B at only .
• Comparing product configurations for different customer segments.
• Van Westendorp (the referenced method) has established the acceptable price range and you need to optimise within it.

Do NOT use when:

• Pricing a single product with no feature variation—Van Westendorp or direct A/B testing (the referenced method) is simpler.
• The product has >6 key attributes—survey complexity increases exponentially with attributes; respondent fatigue degrades data quality.
• The customer base is too small (<100 respondents)—conjoint requires statistical power for reliable part-worth estimation.
• Customer segments are undefined—segment first using Value Proposition Canvas (the referenced method).

Sample Size and Duration

• Respondents: 100–300 per study (50 minimum per segment).
• Design: 4–8 hours (attribute/level selection, survey construction).
• Data collection: 1–2 weeks.
• Analysis: 1–2 days.
• Total: 2–4 weeks end-to-end.

Prerequisites

• Product features defined (from A3 prototype and A4.1 Value Proposition Canvas).
• Acceptable price range established (ideally from Van Westendorp PSM).
• Survey design tool with conjoint capability (Qualtrics, Conjoint.ly, Sawtooth).
• ≥100 respondents from target segment per study (200+ preferred).
• Statistical analysis capability (R, Python, SPSS, or tool-native analysis).
• Time: 2–4 weeks for design, collection, and analysis.

Complete Procedure

Step~1: Define Attributes and Levels (4–8 hours)

Identify 3–6 product attributes (features, service level, brand, price) and 2–4 levels per attribute. Example for C001 Smart Checkout:

Step~2: Generate Choice Sets (2–4 hours)

Use orthogonal design or D-optimal algorithm to generate 8–12 choice tasks, each presenting 2–3 product profiles. Respondents choose their preferred option (Choice-Based Conjoint / CBC is the standard approach).

Step~3: Survey Execution (1–2 weeks)

Present each respondent with the full set of choice tasks. Include demographic questions for segment analysis. Add a “none” option to capture price-outs (respondent would not buy any option).

Step~4: Part-Worth Estimation

Use hierarchical Bayesian estimation (standard in modern conjoint tools) to calculate the part-worth utility of each attribute level. Higher part-worth = stronger preference.

Step~5: Willingness-to-Pay Calculation

Convert part-worths to marginal WTP per feature:

This reveals how much customers will pay per feature—directly informing tier design and feature allocation.

Step~6: Market Simulation

Model different product configurations against each other (including competitor offerings) to estimate market share and revenue. Identify the configuration that maximises revenue or market share, depending on strategic objective.

Quality Criteria

1. Adequate sample: ≥100 respondents per segment.
2. Realistic attributes: Levels reflect actual product options, not hypothetical features.
3. “None” option: Included to capture price-outs.
4. Internal validity: Holdout tasks included to verify prediction accuracy.
5. Segment analysis: Part-worths calculated per segment, not just aggregate.

Theoretical Foundation

Seminal references

• Demonstrated that human decision-making under trade-off pressure reveals preferences more accurately than direct questioning—the theoretical basis for conjoint's forced-choice design.
• Established the experimental design principles (orthogonal arrays, factorial design) that underpin modern conjoint analysis survey construction.

Contemporary references

• Extended experimental design principles to digital product experimentation, showing how conjoint-derived insights guide A/B test design for pricing experiments.
• Positioned conjoint analysis as part of the business model experimentation toolkit, suitable when feature–price interaction effects need quantification.

Challenges and Solutions

Challenge~1: Attribute Overload

Symptoms: >6 attributes produce respondent fatigue; data quality degrades.

Solutions: Limit to 3–6 attributes. Use qualitative pre-study to identify which attributes matter most. Test attributes that are “on the boundary” of pricing tiers (not obviously free or obviously premium).

Challenge~2: Hypothetical Bias

Symptoms: Stated preferences diverge from actual purchase behaviour.

Solutions: Include “none” option. Complement with A/B price testing in A4.3 pilot to validate conjoint-derived pricing in market.

Challenge~3: Context Dependency

Symptoms: Results change based on how attributes are described or framed.

Solutions: Use clear, jargon-free attribute descriptions. Include product demo or video before survey. Pre-test survey with 10–15 respondents for clarity.

Relationship to Other Methods

Conjoint Analysis receives input from:

• Van Westendorp PSM (the referenced method)—acceptable price range constrains price levels.
• Value Proposition Canvas (the referenced method)—feature importance ranking guides attribute selection.
• Customer Development Interviews (the referenced method)—qualitative insights inform attribute framing.

Conjoint Analysis provides input to:

• Unit Economics Modelling (the referenced method)—WTP per feature feeds revenue projections by tier.
• A/B Testing (the referenced method)—conjoint-optimal configurations tested in A4.3 pilot.
• Business Model Canvas (the referenced method)—tier design and pricing update Canvas v2.0.

Example: C001 Smart Checkout — A4.2 Feature–Price Trade-Off

Context: n=180 respondents, Choice-Based Conjoint with 4 attributes × 3 levels = 10 choice tasks.

Key findings:

• 1-click checkout has the highest part-worth (utility 2.1)—customers value speed above all other features.
• Full personalisation adds $2.50 WTP over “none” personalisation.
• Priority support adds only $0.80 WTP—not worth premium tier differentiation.
• Optimal Basic tier: 3-click + basic personalisation at $4.99.
• Optimal Pro tier: 1-click + full personalisation at $7.99.
• Simulated market share: Pro 38%, Basic 45%, None 17%.

Tools and Templates

• Conjoint.ly: Dedicated conjoint tool with survey and analysis.
• Sawtooth Software: Industry-standard conjoint platform.
• Qualtrics: General survey tool with conjoint module.
• R: choiceDes and mlogit packages.
• Python: pyDOE2 for design, custom HB estimation.

• D. J. Bland & A. Osterwalder (2020). Testing Business Ideas. Wiley.
• D. Kahneman (2011). Thinking, Fast and Slow. Farrar, Straus and Giroux.
• R. Kohavi, D. Tang & Y. Xu (2020). Trustworthy Online Controlled Experiments: A Practical Guide to A/B Testing. Cambridge University Press.
• Fisher (1935). fisher1935design.