Conjoint analysis is the scientific way to understand what users truly value — not what they say they want, but what they actually choose when faced with trade-offs.
Conjoint analysis is a powerful method to uncover the features and attributes that actually drive user decisions. It forces respondents to make trade-offs between competing product attributes, revealing what matters most in real choice scenarios. This is critical because what users say they want often differs from what they choose when constrained.
The entire process begins with robust qualitative research to identify the attributes that influence user decisions. From there, you build product profiles combining attribute levels, design a survey to capture preferences, analyze the data statistically, and finally translate the insights into actionable product strategies tailored to distinct user personas.
Start with rigorous user research and attribute definition
The foundation of conjoint analysis is a deep understanding of your users' decision-making process. Your first step is to conduct qualitative research — user interviews, focus groups, or observational studies — to identify the key attributes that users consider when choosing a product.
Talvinder emphasizes:
"Collect detailed transcripts from interviews to capture user preferences, pain points, and thought processes."
Transcribing these sessions enables you to apply qualitative coding techniques — open, axial, and selective coding — to extract recurring themes and attributes. For example, when researching smartphones, attributes might include price, camera quality, battery life, and brand reputation.
After generating a broad list of attributes, you need to narrow it down to the critical few that truly influence decisions. Validate this shortlist with users, stakeholders, and domain experts to ensure relevance.
For each attribute, define levels that cover the realistic range of options users might encounter. These levels must be mutually exclusive and collectively exhaustive. For example, price levels could be ₹10,000, ₹15,000, and ₹20,000.
Product research kickoff at a fintech startup in Bangalore
You (PM): “We need to identify the decision factors that really matter to our users when choosing a savings app.”
Research Lead: “Let's start with interviews and focus groups. We'll transcribe everything and code the transcripts for themes.”
You (PM): “Great. Once we have attributes, we can define levels and construct product profiles for the conjoint survey.”
Getting the right attributes and levels is critical — garbage in, garbage out.
Construct product profiles using experimental design
Once you have your attributes and levels, the next step is to create product profiles — hypothetical product configurations combining different attribute levels.
For example, a project management software profile might specify:
- Number of projects supported: 5
- Number of collaborators: 10
- Price: ₹500/month
- Template availability: Intermediate
You generate multiple such profiles covering various combinations. However, the total number of possible profiles grows exponentially with attributes and levels.
To keep the survey manageable, use experimental design techniques such as fractional factorial designs. These reduce the number of profiles while preserving the ability to estimate utilities accurately.
Talvinder explains:
"Design the survey so respondents can easily compare sets of 3–5 product profiles and choose their preferred option."
This choice-based approach mimics real-world trade-offs and yields more reliable preference data than rating or ranking alone.
Design and implement the conjoint survey carefully
The survey is your data collection instrument and must be designed thoughtfully.
It typically includes:
- Screener questions to filter for your target users
- Conjoint choice tasks where respondents pick among product profiles
- Demographic and psychographic questions for segmentation
Recruit a representative and sufficiently large sample of your target audience. This might mean using online panels, internal customer databases, or external agencies.
Before launching, Talvinder recommends developing personas based on prior research data. Use clustering or factor analysis to identify 4–6 distinct personas that capture meaningful user segments. This sets the stage for segment-level analysis later.
Administer the survey and maintain data quality
Once live, monitor survey response rates and ensure balanced representation across personas.
After data collection, clean the data to remove low-quality responses such as:
- Speeders who finish unrealistically fast
- Straight-liners who give identical answers throughout
Verify that all attribute levels are adequately represented in the dataset to ensure statistical validity.
Calculate utilities and run statistical models
At the heart of conjoint analysis is estimating utilities (also called part-worth values) for each attribute level. These numerical values represent the relative desirability of levels.
Talvinder advises:
"Analyze the survey data using conjoint analysis software like R, Sawtooth, or JMP to calculate part-worth utilities."
Depending on the survey type, apply models such as logistic regression or hierarchical Bayesian models. Validate model fit to ensure robustness.
Interpret utilities carefully:
- High positive utility means strong preference
- Negative utility means aversion or dislike
Use these utilities to analyze trade-offs. For example, determine how much more users are willing to pay for better battery life in a smartphone.
Map utilities to personas for segment-level insights
With personas defined earlier, map the utility values to each persona by calculating average utilities for each attribute level within segments.
This reveals which features resonate with specific user groups.
Create preference profiles per persona that highlight their most and least desired attributes.
Talvinder notes:
"Segment-level simulations predict how changes in product features impact different personas’ preferences."
This allows you to tailor product configurations and messaging for each segment.
Simulate market scenarios and predict share
Using the utilities, run market simulations to estimate how users would choose among competing product profiles.
You can test "what-if" scenarios such as:
- Changing price points
- Adding or removing features
- Bundling options
Simulations estimate potential market share for each profile, guiding portfolio decisions.
Identify key drivers of choice by conducting importance analysis — which attributes most influence decisions.
Product strategy meeting at a SaaS startup
You (PM): “According to simulations, reducing price by ₹500 increases market share by 8%, but adding advanced templates only adds 2%.”
CEO: “Focus on pricing then. The data-driven approach just saved us months of guesswork.”
Prioritizing investment based on simulated impact
Translate insights into strategic product recommendations
Finally, convert your conjoint analysis findings into actionable strategies:
- Develop persona-specific product strategies focusing on features each segment values most
- Optimize your product portfolio by discontinuing unpopular configurations and filling gaps
- Craft targeted messaging and marketing highlighting key attributes for each persona
- Formulate an optimal pricing strategy by analyzing price elasticity and willingness to pay
Compile these into a comprehensive strategic report for stakeholders, illustrating data-driven personas, preference models, and market scenarios.
Validate and iterate continuously
Conjoint analysis is not a one-time exercise.
After launching new product configurations, collect post-launch data to validate predicted preferences.
Refine your utility models and personas with real-world feedback.
Regularly update personas to reflect evolving user behavior and preferences.
Use refined personas as a foundation for future research, user journey mapping, and product development.
- Identify a product category you are familiar with (e.g., food delivery app, smartphone, SaaS tool).
- Conduct 3-5 user interviews or review existing qualitative research to extract key decision-making attributes.
- Define 4-6 attributes and 2-4 levels per attribute.
- Sketch 5 hypothetical product profiles combining attribute levels.
- Draft 3 conjoint choice tasks where respondents pick their preferred profile among 3 options.
- Reflect: How would you recruit participants and segment them into personas?
Test yourself: The Conjoint Analysis Dilemma
You are a PM at a Series A Indian SaaS startup building a project management tool. Your team proposes 10 attributes with 5 levels each for the conjoint survey, which would create over 9 million product profiles. The research lead suggests reducing this using fractional factorial design. The CEO wants to include all attributes to capture every nuance.
The call: How do you balance the CEO's request with practical survey design constraints? What do you communicate to the CEO and the team?
Your reasoning:
You are a PM at a Series A Indian SaaS startup building a project management tool. Your team proposes 10 attributes with 5 levels each for the conjoint survey, which would create over 9 million product profiles. The research lead suggests reducing this using fractional factorial design. The CEO wants to include all attributes to capture every nuance.
Your task: How do you balance the CEO's request with practical survey design constraints? What do you communicate to the CEO and the team?
your reasoning:
Where to go next
- Learn how to conduct effective user interviews: User Research Methods
- Understand how to translate research into product strategy: Product Vision and Strategy
- Explore segmentation and persona development techniques: Persona Development
- Master data analysis and interpretation for product decisions: Metrics and KPIs
- Practice building and evaluating product roadmaps: Roadmapping Fundamentals