This page will summarize the different discrete choice methods used within Sawtooth Software tools. In addition, you may want to review the following resource:
Summary of Available Methods
- Conjoint Value Analysis (CVA) is the original (1970’s) traditional conjoint analysis (full profile, rating or ranking data on single or pairwise concepts)
- Adaptive Conjoint Analysis (ACA) is a computerized, adaptive conjoint analysis technique; very popular in the 80s and 90s
- Choice-based Conjoint (CBC) is currently the most popular technique which asks respondents to choose one of several product concepts (no ratings/rankings)
- Adaptive Choice-based Conjoint (ACBC) uses a build-your-own screening section together with a choice task tournament to identify products that are most likely to be first considered by respondent and then chosen
- Maximum Difference Item Scaling (MaxDiff) is roughly comparable to a one-attribute, multi-level CBC where respondents typically shown 2-6 items at a time, asked to indicate which is best and which is worst, and then the task is repeated many times showing a different set of items in each task
- Menu-based Choice (MBC) is an analysis tool that allows researchers to model complex decisions, such as restaurant menus, multi-drug prescriptions, telecom bundling, etc.
Which Conjoint Method Grid
| Method | Minimum Sample Size | Attributes | Levels | Pricing? | Complexity* | Fielding | Typical Use |
|---|---|---|---|---|---|---|---|
| CVA | Small | Typical: 6-7 Max: 30 |
Typical: 4-5 Max: 15 |
Yes, but limited | No | Paper, computer | Small attribute studies, situations where objective is to measure purchase likelihood or other direct scale elicitation, small sample size studies; may be used to generate generalized designs; situations where small, fixed design is required. |
| ACA | Small | Typical: 7-8 Max: 10 |
Typical: 5 Max: 15 |
Not recommended | No | Computer only | Large attribute studies; situations where objective is to measure purchase likelihood. |
| ACA 30 | Small | Typical: 12 Max: 30 |
Typical: 5 Max: 15 |
Not recommended | No | Computer only | Large attribute studies; situations where objective is to measure purchase likelihood. |
| CBC | Large | Typical: 6-7 Max: 10 |
Typical: 5 Max: 15 |
Yes | Yes | Paper, computer |
Competitive scenarios where choice is among multiple alternatives; pricing studies; chip allocation studies; fixed alternatives/competitors. |
| CBC ADM | Large | Typical: 6-7 Max: 250 |
Typical: 5 Max: 250 |
Yes | Yes | Paper, computer |
In addition to the CBC usage above, the Advanced Design Module allows you to conduct alternative-specific studies, partial-profile designs, and shelf-facing studies with up to 100 product concepts per screen. |
| ACBC | Small | Typical: Any Max: 100 |
Typical: Any Max: 250 |
Yes | Some | Computer only | Pricing studies; large number of attributes; focus is on finding best product; allow respondents or situation to determine which attributes/levels are shown. |
| MaxDiff 30 | Medium | No Attributes | Typical: 20-25 Max: 30 |
N/A | No | Paper, computer |
Lists of brands, positioning statements, specific product concepts, flavors, etc. |
| MaxDiff 2000 | Medium | No Attributes | Typical: 30-40 Max: 2000 |
N/A | No | Paper, computer |
Lists of brands, positioning statements, specific product concepts, flavors, etc. |
| MBC | Very Large | Typical: Any Max: 1000 |
Typical: Any Max: 1000 |
Yes | Yes | Paper, computer |
Multi-part decisions; complex models; bundling; mixed designs (CVA & CBC together). |
*Complexity, in this case, means whether or not the attributes can freely combine with other attributes.
Strengths and Weaknesses of the Different Methods
Conjoint Value Analysis (CVA)
- [Strength] Good for both product design and pricing issues
- [Strength] Can be administered on paper, computer/internet
- [Strength] Shows products in full-profile, which many argue mimics real-world
- [Weakness] Limited ability to study many attributes (more than about six results in respondent fatigue, long questionnaires)
- [Weakness] Limited ability to measure interactions and other higher-order effects (cross-effects)
Adaptive Conjoint Analysis (ACA)
- [Strength] Ability to measure many attributes, without wearing out respondent
- [Strength] Respondents find interview more interesting and engaging
- [Strength] Efficient interview: high ratio of information gained per respondent effort
- [Weakness] Partial-profile presentation less realistic than real world (respondents may not be able to assume attributes not shown are “held constant”)
- [Weakness] Often not good at pricing research (tends to understate importance of price, and within each respondent assumes all brands have equal price elasticities)
- [Weakness] Must be computer-administered (PC or Web)
Choice-Based Conjoint (CBC)
- [Strength] Easy for respondents to answer
- [Strength] Flexible designs (full or partial profile designs, alternative-specific designs including multiple fixed alternatives, chip allocation, shelf-facing designs, paper & pencil or computer administered)
- [Strength] Measure “None” alternative
- [Strength] Works well for measuring price
- [Weakness] Best to have large sample sizes (n=>200). If your sample size is smaller, you’ll need to keep your attribute list small and ask each respondent to answer more CBC tasks.
- [Weakness] Limited number of attributes – survey can become too cumbersome for respondents (but you can use partial profile designs)
- [Weakness] Not adaptive
Adaptive Choice-Based Conjoint (ACBC)
- [Strength] Many of benefits of CBC, but can be done with smaller sample size
- [Strength] Good choice if about 5 or more attributes
- [Strength] Works well for measuring price
- [Strength] Accommodates non-compensatory behavior
- [Strength] More attributes
- [Strength] More interaction leads to greater respondent involvement in survey
- [Weakness] Survey is often 2-3 times longer than a comparable CBC
- [Weakness] Currently no support for some CBC “goodies” (chip allocation, traditional none, etc.)
- [Weakness] More complex to program, analyze
- [Weakness] Must be administered on computer
- [Weakness] May be overkill for small-attribute studies (4 or fewer attributes)
Maximum Difference Item Scaling (MaxDiff)
- [Strength] Easy for respondents to answer
- [Strength] Fixed designs possible (supports paper-and-pencil interviewing)
- [Strength] Often better than standard rating or ranking exercises (ratings often end up in ties; rankings are difficult to manage with more than about 7 items)
- [Weakness] Best to have large sample sizes (n=>200)
- [Weakness] Resulting model is not additive (can’t add the score for one item to the score for another item to find the value of offering both)