Method
Conjoint analysis: which features do your customers buy – and at what price?
In a conjoint analysis, respondents repeatedly choose between offers that differ in a few attributes. From their choices, 121surveys calculates how important each attribute is, what each level is worth and which offers would win in the market.
Building and testing is free. To publish, you choose a paid plan.
Example: what matters in a streaming subscription
Respondents repeatedly choose between subscription offers that differ in picture quality, ads, number of devices and price. From their choices, the analysis works out what drives the decision.
How strongly each attribute drives the choice
Importance: share of the range of all part-worths.
- Picture quality
- 13%
- Ads
- 27%
- Devices at once
- 19%
- Price per month
- 42%
What a level is worth
Price premium per month over the weakest level of the same attribute, estimated from the price attribute.
| Level | Premium |
|---|---|
| Picture quality | |
| HD | Baseline |
| 4K | +€2.38 |
| Ads | |
| with ads | Baseline |
| ad-free | +€5.14 |
| Devices at once | |
| 1 device | Baseline |
| 2 devices | +€1.94 |
| 4 devices | +€3.52 |
Simulated responses from 400 people with 10 tasks each, calculated with the 121surveys analysis (logit). As of 11/9/2026.
Conjoint analysis explained
What is a conjoint analysis for?
Finds out which product features really drive the choice. Instead of asking “What matters to you?” directly – where everything ends up mattering – people choose between realistic offers.
What do respondents see?
Several rounds of two or three product offers side by side, for example with different brands, sizes and prices, and the question of which one they'd pick.
How many responses do you need?
Published studies usually work with 100 to 300 respondents. As a rule of thumb, aim for 300 responses and 200 per group you want to compare.
How do you read the result?
Importance shows how strongly a feature shapes the choice – the longer the bar, the more decisive it is. Part-worth utilities show which level within a feature is popular (positive) and which puts people off (negative); only compare them within the same feature. In the simulator, you put together offers and see how choices would be split.
Where are the limits?
More than about six features overwhelm respondents. The shares in the simulator are choice shares within the survey, not a market share forecast – brand awareness and availability aren't included.
What 121surveys can do for conjoint
Full or partial profiles
Show all attributes or only a selection per task – with many attributes, each task stays readable. Individual attributes can be shown in every task.
Several response formats
Choose one offer, then “would you buy it?”, best and worst offer, allocate points, a paired comparison on a scale, or adaptive with a build-your-own step, screening and tournament.
Logit and Hierarchical Bayes
The analysis estimates part-worths for the whole sample (logit) and, if you want, for every single person (Hierarchical Bayes, runs in the background).
Willingness to pay
If the design includes a price attribute, 121surveys calculates how much of a price premium each level is worth.
Market simulator and portfolio
Put offers up against competitors and see the shares of choice. Portfolio optimisation searches for the best product combinations by reach or contribution margin.
Display for respondents
Cards, table, rows, carousel or shelf, with explanations per attribute and its own attribute order for every person.
How a conjoint study works
Define the attributes
Decide how the offers differ – for example price, performance or contract length. The method guide tells you how many responses your design needs.
Let respondents choose
121surveys puts together balanced tasks for every person. You share the survey via link, QR code or invitation.
Analyse and simulate
Read importances, part-worths and willingness to pay, compare offers in the simulator and export the result as a report or presentation.
Frequently asked questions
What is a conjoint analysis?
A market research method: respondents choose between offers made up of attributes and levels. A statistical model then estimates how much each level contributes to the choice (part-worth) and how important each attribute is.
How many respondents do I need?
It depends on the attributes, levels and number of tasks. The method guide in the editor calculates a minimum number for your design.
What is the difference between logit and Hierarchical Bayes?
Logit estimates one part-worth per level for all respondents together and is available instantly. Hierarchical Bayes also estimates it for each person; it takes longer and shows how differently groups decide.
Can I test prices?
Yes: price becomes an attribute with several levels. From it, 121surveys calculates the willingness to pay per level, and in the simulator you see how shares of choice shift with price.
How much does a conjoint study with 121surveys cost?
Creating and testing a design is free. To publish, you need a paid plan – every plan includes conjoint. What responses cost is shown on the pricing page.
Start your conjoint study
Creating a design and running through it with the test link is free.
