Statistical techniques can often be quite daunting. Most recently the quantitative team had a look at the key driver analysis, a powerful tool aimed at identifying the brand attributes that are most influential on physicians' decision to prescribe. Vicky McLellan continues our campaign to de-mystify these. This month looking at conjoint techniques.
Vicky McLellan 
18 Nov, 2008
Quant Corner: Insight into... Conjoint techniques
In all walks of life all choices that consumers are confronted with require some degree of comprise and trade off, by replicating this decision making process conjoint is a powerful tool that helps build and develop marketing strategies.
Similarly to key driver analysis conjoint looks to determine the relative
importance of brand attributes; but here, attributes are considered jointly.
That is to say that a conjoint involves a choosing task where the respondent
is asked to make trade-offs between different attributes of hypothetical products.
Presenting respondents with different purchasing scenarios unearths the hidden
drivers and non drivers behind product choice.
A key benefit of conducting a conjoint task is the main output of the conjoint analysis, the market model. This is an excel simulator that enables the end user to model and test different ‘what if’ scenarios by selecting various combinations of the product features. From this, likely market preferences and potential market shares can be evaluated based on what customers value the most.
Understanding what values are placed on different product features will:
- Determine the elements that make up the optimal product profile to drive future R&D and clinical trial strategy
- Identify how a market is segmented in terms of the benefits prescribers may seek to enable effective targeting
- Determine how well a new product will stand up against its competitors and guide strategies as to how best the product can defend itself
Different studies, different conjoints
There are a number of different conjoint techniques that can be utilised, depending on the requirements of the research. Although, over the years the two most frequently used techniques are the Choice Based Conjoint (CBC) and the Adaptive Conjoint Analysis (ACA).
Although CBC and ACA are based on the same guiding principles, of these two techniques the choice-based conjoint is overwhelmingly the most common option and is academically favoured for its rigour. This is mostly because of its design, implementation and analysis output. The main advantages of conducting a CBC over an ACA is that a CBC:
- Uses a full product profile that shows all product attributes
- Produces scenarios that replicate real market choices
- Allows the product to be evaluated within a competitive set
- Provides a direct measure of preference share
- Allows the end user to conduct a number of what if scenarios which can be used to estimate relative sales and profitability
This is not to say that an ACA should be automatically discounted as an option. An ACA is often the most suitable approach if an early exploration of attributes is required, when typically the attribute list is quite lengthy. To some degree an ACA can also help make go/ no go decision as an ACA, like the CBC, will give you preference shares. However, where it falls down is that it is difficult to convert these preference shares into volume predictors. This is because an ACA will adapt the choices on an individual level in line with the attributes that the respondent considers to be most important. Thus, the attributes that are used for the trade off will vary on an individual level and the decisions made in this environment are not based on the whole picture.
There is, however, one warning should be heeded with all conjoint techniques: any stated preference shares are the estimated share of prescribing by HCPs and assume 100% awareness and 100% availability. Preference shares tend to overstated and represent the maximum share so a deflation is required.
Deciding on which conjoint technique is best suited to the task in hand is only one aspect that needs to be considered when designing a conjoint study. It is also important to give thought to the level of analysis that is required. To get the most out of a conjoint analysis it is possible to drill down to understand how the perceived value of different products varies by different subgroups of customers. This enables more effective targeting of customers by guiding marketing strategies on a sub group level. Marketing activities can then focus on the product features that are of most value to the individual sub groups of customers, which are likely to be different. To do this a ‘needs based’ segmentation groups respondents in accordance to their preference for different attributes and levels. This analysis is conducted alongside the modelling process and provides a robust segmentation from which decisions can be made. Although it is important to bear in mind that robust sample sizes are required to allow this natural segmentation to occur.
It is also possible within a conjoint analysis to understand how the relative preference of product features varies by different patient types. To do this respondents are asked to complete the conjoint in context of different patient types. This is extremely useful if marketing strategies are aimed at targeting different types of patients. Although it is important to note that this makes the design of the conjoint more challenging as there is a need to balance the scenarios across the different patient types. This could have cost and timing implications.
Within this article we have focused on providing a short overview of the two most popular conjoint techniques, CBC and ACA, and the type of sub-analysis that can be conducted within the remit of these two techniques. There are a number of other conjoint techniques available for consideration.
However when designing any conjoint the following top 7 tips should be taken into consideration.
Top Tips!
- Give due consideration as to what the outputs will be used for, this will affect the design of the conjoint
- Attributes should be mutually exclusive and not overlap
- The levels within the attributes should be comprehensive and relevant to the product and the market
- The number of levels for each attribute should be limited to 4-5 but this will depend on the number of attributes that need to be tested
- Do not test too many attributes; results are likely to be diluted and less reliable.
- Unless the product is in very early development a clear idea of the product features should be known
- Ideally a conjoint should look to test up to 8 attributes
- Conduct a short round of qualitative research or an extended pilot phase to make sure that the attributes meet all of the above requirements!
- Remember stimuli needs to be shown which rules out CATI or other forms of telephone interviews
- Minimum sample size for a conjoint task is 50 per respondent group, this is considered a small sample size
- Small samples may limit the analysis to “Aggregate” results only and might not provide enough weight to assess small effects
Your response:
Login to respond to this article: