Choice-Based Conjoint Analysis: Building Better Market Research [Step-by-Step Guide]

What is Choice-Based Conjoint Analysis and Why It Matters

Choice-Based Conjoint (CBC) analysis is the most accessible survey-based approach to learn about and predict how people make choices when facing challenging tradeoffs. CBC has become the leading conjoint-related technique in the past three decades. This method gives great insights by copying how people make decisions in real life, which helps businesses develop products, set prices, and create market strategies.

How CBC is different from traditional surveys

Traditional surveys ask respondents to rate or rank product concepts one by one. CBC takes a different approach by showing respondents complete product profiles and asking them to pick what they'd likely buy. This creates several advantages.

CBC better shows how consumers actually shop. People don't rate alternatives based on their priorities when shopping—they just pick one product over others. This makes CBC data better at predicting actual market behavior.

It also lets researchers include a "None" option, just like in real life where buyers don't have to choose products they don't want. Researchers can figure out how many people wouldn't pick any of the options shown, which helps project market share.

CBC's method lets them assess how attributes work together. Unlike older conjoint methods that assume attributes work independently, CBC shows how brand and price might affect each other, or why certain colors work better with specific styles. This gives deeper insights into consumer decisions.

Discrete choice vs. choice-based conjoint analysis

People often use "discrete choice" and "choice-based conjoint" to mean the same thing, but there are subtle differences in how they're used and their history. Discrete choice experiments (DCE) or discrete choice modeling (DCM) started in economics and transportation research, while CBC came from marketing research.

CBC is really just a specific type of discrete choice modeling used for marketing. Both methods share the same basic ideas: they show people multiple product concepts at once and ask them to choose rather than rate. They're both based on random utility theory, which says consumers pick options they see as most valuable.

The main difference lies in how they're used. Discrete choice modeling might look at transportation choices or public policy decisions, while CBC focuses on commercial products, pricing strategies, and market segments.

CBC analyzes how people prefer products when compared to competitors, which helps understand how new products might do against existing ones.

Real-life decision simulation in CBC

CBC's biggest strength is how well it copies actual purchasing decisions. A well-designed CBC survey creates scenarios that match the trade-offs consumers face while shopping.

A typical CBC exercise asks respondents to look at about ten choice tasks. Each task shows three to six product concepts with different attributes. Every concept includes one level from each attribute (like brand, price, features), so researchers can see which attributes matter most.

By analyzing these choices statistically, researchers can calculate "part-worth utility scores" for each attribute level. These scores show how much consumers value specific product features. Attributes with bigger utility ranges play a larger role in decision-making.

These utilities power market simulators that predict consumer response to many product configurations without testing each one. Companies can try different product designs, pricing strategies, and competitive scenarios to estimate market acceptance, price sensitivity, and potential cannibalization effects.

While CBC can't account for everything that affects market shares (like distribution or advertising), it gives accurate relative preference indicators when designed properly. This makes it the foundation for optimizing products and predicting market performance before investing heavily in production and marketing.

Designing a Choice-Based Conjoint Survey Step-by-Step

Creating a choice-based conjoint survey that works needs careful planning and execution. This guide will help you design a survey step by step to get meaningful results.

Selecting attributes and levels

The success of any conjoint study depends on choosing the right attributes and levels. You should pick attributes that stand completely independent of each other. Using attributes with overlapping meanings wastes time and creates biased utility results.

Your selected levels for each attribute must be:

• Mutually exclusive within each attribute
• Cover the full range of possibilities for existing and potential products
• Limited to about 5 or fewer levels per attribute whenever possible

The number of levels you pick can affect your results by a lot due to the "Number-of-Levels Effect." Attributes with more levels tend to become more important in the analysis. For measurable attributes like price or speed, stick to five to seven levels to avoid data reversals.

Choosing the right choice task format

Choice tasks come in several formats. Single-choice remains the most common option where people pick what they prefer from a set. This approach matches real-life buying decisions most closely.

Best-worst conjoint could work better as respondents mark both their most and least preferred options. Your data gets more variance this way. Products that might involve multiple purchases in one shopping trip work well with a continuous sum approach.

Setting number of profiles and sets per respondent

Full-profile presentation works best when concepts stick to eight attributes or less to keep respondents focused. Choice tasks usually work well with 3-5 alternatives.

Most CBC questionnaires include 8 to 15 questions per respondent. Web surveys might need fewer tasks if you can get more respondents. People using hierarchical Bayes estimation need about 10 choice tasks or more for solid individual-level predictions.

Avoiding prohibited level pairs

You shouldn't use prohibited pairs much of either. Some attribute level combinations just don't make sense together. To name just one example, testing in-home tech might need to exclude Amazon Echo devices with Google Assistant operating systems.

These alternatives work better than creating prohibitions:

• Tell respondents they'll see combinations not yet in the market
• Combine related attributes into a single attribute with more levels
• Use conditional pricing or graphics for certain combinations

Too many prohibitions can mess up your effects and make it impossible to calculate stable utilities.

Sample size recommendations for segmentation

Conjoint studies typically need 150-1,200 respondents. Segmentation analysis works best with at least 200 respondents per segment. A four-segment study needs about 800 respondents minimum.

Solid quantitative research without segmentation needs at least 300 respondents. Research by Johnson and Orme shows that ten tasks per respondent roughly equals having ten times as many respondents do one task.

Want to create your first CBC survey? SurveySparrow's easy-to-use tools help build choice-based conjoint surveys that show what customers want and how to make your products better.

Running the Survey and Collecting Data

After designing your choice-based conjoint survey, proper execution and data collection need careful planning. The statistical validity of your results and respondent experience will improve with good implementation.

Randomization of choice sets

Randomization of choice sets is vital in CBC studies. Modern CBC implementation assigns different questionnaire versions randomly to participants instead of giving everyone similar questionnaires. This approach provides many benefits.

The order of tasks helps control context effects across respondents when randomized. CBC's design algorithms create concept combinations with good statistical properties automatically. These tasks aren't truly random despite being called "random choice tasks".

Several randomization strategies are available:

• Complete Enumeration: Creates profiles that are nearly orthogonal within respondents, with equally balanced frequencies between attributes. This approach minimizes overlap of attribute levels within choice sets.
• Shortcut: Constructs profiles using the least frequently used attribute levels for each respondent, maintaining minimal overlap within tasks.
• Random: Samples profiles randomly from all possible combinations, allowing some overlap to occur.
• Balanced Overlap: Offers a compromise between Complete Enumeration and Random, allowing moderate overlap.

Multiple versions of the questionnaire should be included before launching your study. Single-version plans can produce good results, but adding a few more versions will improve your data quality by reducing psychological effects.

Ensuring balanced and orthogonal design

A balanced design happens when each level within an attribute appears equally throughout your survey. At the same time, orthogonality means designs have zero correlation between pairs of attributes. Both properties help drive design efficiency.

CBC's approach produces exceptional designs automatically, though perfectly orthogonal and balanced designs might be hard with a reasonable number of tasks. The software creates a pool of potential conjoint questions through algorithms that select efficiently balanced designs based on your specifications.

CBC uses D-efficiency calculations to review the design's ability to estimate part-worth utilities. These measurements help you check if your design has the right statistical properties before collecting data.

Mobile and desktop compatibility considerations

Survey optimization for mobile devices has become essential due to increased smartphone usage. Well-formatted mobile surveys can boost completion rates and data quality by a lot.

Matrix tables, multiple choice, slider, and side-by-side questions need special attention for mobile optimization. Here are some examples:

• Matrix tables: Mobile devices can show these in accordion format when tables are too large.
• Multiple choice: Vertical list format works better than columns on mobile.
• Sliders: "Mobile friendly" setting ensures they fit on screen without scrolling.

Your survey should be tested on different device types to ensure it works well. Most platforms let you preview how your survey looks on various devices.

Ready to run your choice-based conjoint study? SurveySparrow's strong mobile-optimized survey tools will help your CBC analysis capture quality data across all devices.

Analyzing Results and Calculating Part-Worth Utilities

You need to analyze the results to find meaningful insights after collecting choice data from your survey. The statistical analysis of choice-based conjoint data helps calculate utilities. These utilities show how each attribute level affects consumer choice.

Understanding part-worth utility scores

Part-worth utilities are numbers that show how features influence customer's choices. These scores show how much people want each attribute level. Higher scores mean people prefer those features more.

You should know that utilities are interval data scaled to an arbitrary constant within each attribute. This lets you compare values within the same attribute, but not between different ones. Let's look at an example. "Brand A" with a utility of 20 and "Brand B" with 40 shows a meaningful preference increase of 20 points. But you can't compare "Red" (utility 20) to "Brand A" (utility 20) and say people want them equally.

Most modern conjoint analysis software makes utilities zero-centered within each attribute. This scaling method makes attributes add up to zero, which can be confusing. A negative utility value doesn't mean people dislike that level. It just shows they preferred other levels in that attribute more.

Attribute importance calculation

The relative importance of each attribute depends on its potential impact on a product's total utility. We calculate this by finding the range between highest and lowest utility values within each attribute.

Here's an example:

Brand (B - C): 60 - 20 = 40    Importance: 26.7%

Color (Red - Pink): 20 - 0 = 20    Importance: 13.3% 

Price ($50 - $100): 90 - 0 = 90    Importance: 60.0%

Total: 150    Total: 100% 

Note that importance scores depend on the attribute levels you choose for your study. Price would seem less important with a smaller price range. You should calculate importances for each respondent first, then average them. Don't compute importances from average utilities.

Using hierarchical Bayes for individual-level estimates

Hierarchical Bayes (HB) stands as the best method to estimate individual-level utilities in choice-based conjoint studies. HB shows unique preference patterns among respondents, unlike aggregate logit models.

HB gives each person in your study a separate utility. The overall utilities are just averages of individual utilities. This helps you learn more about different segments and makes market simulators more accurate.

The process works by calculating possible value distributions for each respondent's attribute level. These possible utilities are "draws." The average of these draws gives you the best estimate of what someone truly prefers.

HB estimation can help you discover powerful individual-level insights that aggregate models can't provide. This makes it valuable for your choice-based conjoint analysis.

Using CBC Insights for Product and Pricing Decisions

CBC analysis shows its true value when businesses apply its insights to crucial decisions. The findings from preference data can transform your business strategy.

Market share estimation from choice based conjoint

CBC gives "shares of preference" that show predicted product shares, assuming equal awareness and distribution—not actual market shares. Several effective approaches can turn these into revenue projections:

• Results tracked over time establish a "discount factor" between predicted and actual shares
• Historical actions modeled and compared to actual changes
• A conservative approach cuts predicted share increases by half

Running simulations with different product profiles

Market simulators stand out as CBC's most valuable output. These tools help you test countless product configurations without building physical prototypes. Different scenarios can predict how markets react to:

• Feature additions or removals
• Competitive product introductions
• Various positioning strategies

Testing price sensitivity and revenue projections

Price sensitivity analysis shows the best price points to maximize profit. The scale affects projected equilibrium prices substantially—higher scale results in lower equilibrium prices. The elasticity calculation works like this:

"The price elasticity of demand measures how sensitive quantity demanded is to price changes. For example, if a 10% price increase leads to a 20% quantity decrease, elasticity would be -2."

Conjoint analysis examples from real businesses

Major companies have used CBC to make important business decisions.

Microsoft found optimal pricing for peripheral improvements and product configurations. Procter & Gamble's CBC predictions matched actual market price sensitivity closely. A storage media manufacturer's yearly conjoint studies helped set price premiums against competitors. Lifetime Products used CBC data to convince a retailer about higher prices—which ended up increasing profits.

Conclusion

Choice-Based Conjoint analysis is a powerful tool that reshapes our understanding of customer priorities and decision-making processes. This piece shows how CBC mirrors real-life purchasing scenarios and delivers more accurate predictions than traditional survey methods.

The trip from survey design to result analysis might seem complex initially. The payoff makes this effort worthwhile. CBC provides solid data about features that matter most to customers, their price sensitivity, and trade-offs they accept.

Market simulation capabilities let you test countless product configurations without expensive prototypes or risky launches. You can predict customer responses to new offerings, competitive entries, or price changes before making major investments.

Companies of all types have seen the benefits of this approach. Their success stories show how CBC insights lead to better product development, smarter pricing strategies, and higher profits.