Tuhin Bhuyan · 19 January 2026 · 9 min read
Conjoint analysis reveals what users truly value by making them choose between realistic options. This guide covers study design, utilities, willingness-to-pay, and how to turn trade-off data into clearer product and pricing decisions.
Conjoint analysis is a survey-based research method that measures how people value different attributes of a product or service.
Instead of asking "how important is feature X?" (which produces vague, inflated answers), it presents respondents with complete product configurations and asks them to choose.
The choices reveal what actually matters.
The method was developed by marketing professor Paul Green at Wharton in the 1970s and has since become one of the most widely used techniques in market research.
The reason it stuck: it mirrors how real decisions work. Nobody evaluates features in isolation. They weigh trade-offs. More storage or lower price?
Better design or faster performance?
Conjoint analysis forces those trade-offs in a controlled setting, then uses statistical analysis to decompose the choices into measurable preference values.
The output is a set of numbers called part-worth utilities. Each number tells you how much a specific feature level contributes to overall preference.
High utility means people want it. Low utility means they don't.
The spread between the highest and lowest utility within an attribute tells you how much that attribute matters relative to others.
That's the practical value. You stop debating which features to include in your next release based on opinions and start making those decisions based on measured preference data.
Direct preference questions are popular because they're easy to write.
"On a scale of 1 to 5, how important is battery life?" The problem is that everyone rates everything as important.
You end up with a flat list of 4s and 5s that tells you nothing about priority.
Real purchasing decisions don't work that way. When someone picks a laptop, they're not rating battery life in a vacuum.
They're weighing battery life against price against weight against screen quality. Something has to give. The laptop with 15-hour battery life costs more.
The lightweight one has a smaller screen. Every choice involves a sacrifice.
Conjoint analysis replicates this.
Would you prefer a laptop with 12-hour battery and 256GB storage at $1,100, or one with 8-hour battery and 512GB storage at $900?
Now you're measuring how much battery life is worth relative to storage and price.
Not in the abstract, but in the context of a real decision.
This approach surfaces preferences that respondents themselves may not consciously recognise. People often can't articulate why they chose one option over another.
But their pattern of choices, across multiple trade-off scenarios, reveals a consistent value structure that you can quantify and act on.
["IMAGE - Side-by-side comparison showing a direct survey question ('Rate the importance of battery life: 1-5') producing flat, uninformative results versus a conjoint choice task showing two product profiles with varying attributes (battery, storage, price) that forces a meaningful trade-off decision."] ["ALT - Side-by-side comparison showing a direct survey question ('Rate the importance of battery life: 1-5') producing flat, uninformative results versus a conjoint choice task showing two product profiles with varying attributes (battery, storage, price) that forces a meaningful trade-off decision."]
Not all conjoint studies work the same way. The two most common approaches are Choice-Based Conjoint and Adaptive Conjoint Analysis.
Which one you use depends on how many attributes you're testing and how much precision you need.
CBC is the most widely used form of conjoint analysis today.
It presents respondents with sets of 2 to 4 product profiles and asks them to pick their preferred option.
This mimics real shopping behaviour: you look at a few options and pick one.
A typical CBC study shows 8 to 15 choice tasks.
Each task presents a different combination of attribute levels, and the experimental design ensures that every attribute level appears enough times for reliable statistical estimation.
The "none of these" option is often included so respondents can opt out, which produces more realistic demand estimates.
CBC works well for studies with 4 to 7 attributes. Beyond that, the choice tasks become cognitively demanding and response quality drops.
ACA adjusts the questions as the respondent answers.
Early questions identify which attributes matter most to that individual, and later questions focus on the trade-offs within those attributes.
This makes ACA more efficient for studies with many attributes (8 or more) because respondents only see trade-offs that are relevant to their preferences.
The trade-off is complexity. ACA requires more sophisticated survey logic and larger sample sizes to produce stable estimates. For most product teams, CBC covers the ground you need.
Before designing a conjoint study, you need to understand four terms. They're not complicated, but getting them right determines whether your study produces actionable data or noise.
There's also relative importance, which is derived from part-worth utilities.
If price utilities span a range of 40 points and storage utilities span a range of 15 points, price is roughly 2.5 times more influential in driving choices.
This gives you a clean percentage breakdown of what matters most.
["IMAGE - Diagram showing the relationship between conjoint analysis concepts: attributes (price, storage, support) each containing levels (e.g. $29/$49/$99 for price), combining into profiles (complete product configurations), which respondents compare in choice tasks. Arrows show how choices are decomposed into part-worth utilities and relative importance percentages."] ["ALT - Diagram showing the relationship between conjoint analysis concepts: attributes (price, storage, support) each containing levels (e.g. $29/$49/$99 for price), combining into profiles (complete product configurations), which respondents compare in choice tasks. Arrows show how choices are decomposed into part-worth utilities and relative importance percentages."]
A well-designed conjoint study starts with clear decisions about what you're testing and who you're asking. Get these wrong and no amount of statistical analysis will save you.
Sample size depends on your design complexity.
For a basic CBC study with 5 attributes and 3 levels each, 200 to 300 respondents typically produce stable estimates.
If you plan to segment the data (by company size, role, or geography), you need enough respondents in each segment.
Plan for 150 or more per segment.
Conjoint analysis produces several outputs. Each one answers a different question about how your customers make decisions.
This tells you what percentage of the overall decision each attribute accounts for.
If price explains 35% of preference, brand explains 25%, and features explain 20%, you know that pricing strategy matters more than feature additions for this audience.
These percentages add up to 100% and give you a clear hierarchy of what drives choices.
Each attribute level gets a utility score. Within the price attribute, $29/month might score +15, $49/month might score +3, and $99/month might score -18.
The pattern is obvious: lower prices are preferred.
But the gap between $29 and $49 (12 points) is much smaller than the gap between $49 and $99 (21 points).
That tells you the jump to $99 is where you lose people, not the jump to $49.
Look for surprises. Sometimes a feature level that seems minor drives significant preference. Sometimes a "premium" option doesn't justify its price premium. These non-obvious findings are where conjoint analysis earns its value.
By comparing feature utilities to price utilities, you can calculate how much customers would pay for specific improvements.
If adding a dedicated account manager increases utility by 8 points, and each dollar of monthly price reduces utility by 0.5 points, that feature is worth roughly $16/month to your customers.
This turns abstract preference data into concrete pricing decisions.
Once you have utility scores, you can model hypothetical scenarios. What happens to your market share if you add a feature?
What if a competitor drops their price?
Market simulations let you test product and pricing changes before committing resources, which is considerably cheaper than launching something and hoping for the best.
["IMAGE - Example conjoint analysis output showing three panels: (1) a horizontal bar chart of relative importance percentages for five attributes, (2) a table of part-worth utilities for each attribute level with the highest utility highlighted in each row, and (3) a willingness-to-pay calculation showing how feature utilities convert to dollar values."] ["ALT - Example conjoint analysis output showing three panels: (1) a horizontal bar chart of relative importance percentages for five attributes, (2) a table of part-worth utilities for each attribute level with the highest utility highlighted in each row, and (3) a willingness-to-pay calculation showing how feature utilities convert to dollar values."]
Conjoint analysis answers a specific question: what do people value when they're forced to choose between product configurations? It's a trade-off measurement tool.
Use it when you need to understand how features interact with each other and with price.
Good fits for conjoint analysis:
Where conjoint falls short: it doesn't measure satisfaction with existing features (that's what Kano analysis is for).
It doesn't find your optimal price point in isolation (that's Van Westendorp ). And it requires more setup than a simple poll.
If you just need a quick read on a single question, a FastPoll is faster and simpler.
Conjoint is most valuable when the decision space is complex: multiple attributes, multiple levels, and real trade-offs between them.
If you're choosing between two options on a single dimension, you don't need conjoint.
If you're designing a product with five configurable dimensions and three price tiers, you do.
Conjoint studies are powerful when designed well and misleading when designed poorly. These are the errors that produce data you can't trust:
Running a conjoint study used to require specialised research software, a statistics background, and weeks of analysis time.
The survey design alone (creating balanced experimental designs that ensure every attribute level appears with sufficient frequency) was a project in itself.
Most product teams either hired a research firm or skipped conjoint entirely.
The survey part is manageable. You define attributes and levels, create choice tasks, and collect responses. The analysis is where things get difficult.
Estimating part-worth utilities requires hierarchical Bayesian estimation or multinomial logit models.
Calculating willingness to pay means converting utility differences into dollar values using the price attribute as a reference.
Running market simulations means building choice probability models. None of this is something you want to do in a spreadsheet.
SenseFolks UserChoice handles the entire conjoint workflow.
You define your attributes and levels, the platform generates a statistically balanced experimental design, and the analysis runs automatically as responses come in.
No manual utility estimation. No simulation modelling. You get part-worth utilities, relative importance scores, willingness-to-pay calculations, and segment breakdowns on your insights dashboard .
Here's how to set it up:
Because SenseFolks follows the Website → Survey → Insights model, your conjoint data lives alongside your other product research.
Pricing studies, feature prioritisation surveys, content reaction tests: everything aggregates into one dashboard per website.
Product configuration decisions are better when they're informed by the full picture, not just one data source.
Every product decision is a bet. You're betting that customers want this combination of features at this price.
Conjoint analysis doesn't eliminate the bet, but it tells you the odds before you place it.
The teams that consistently ship products people actually want aren't luckier than everyone else. They just measure preference before they commit resources.
["IMAGE - SenseFolks UserChoice survey results screen showing conjoint analysis output with relative importance bars for each attribute, part-worth utility scores for each level, and a willingness-to-pay summary on the insights dashboard."] ["ALT - SenseFolks UserChoice survey results screen showing conjoint analysis output with relative importance bars for each attribute, part-worth utility scores for each level, and a willingness-to-pay summary on the insights dashboard."]
Create a UserChoice survey and get trade-off outputs directly in your insights dashboard.
