Conjoint analysis – the easy way to compare apples with pears

Conjoint analysis seems to have been around for Eons – well at least since the 1960s when it originated within Mathematical Psychology.

So why is such a valuable tool regarded as the strict preserve of high-end management consultants and a few boffins at select research agencies?

At Nepa we believe this is chiefly down to lack of accessibility and the way in which the technique is portrayed as some kind of mystical ‘black box’. Coupled with the perceived high associated development and execution costs, conjoint seems to be unfairly relegated to the ‘stiffs’ on the bench at most research companies – even those in the premier league.

We argue that it is time for a new look at conjoint to bring it back into the research mainstream.

The key premise of conjoint analysis is that respondents are generally unable to give a correct weight to a product or service’s individual attributes or features. This can be overcome when they are asked to consider these as a bundle of multiple conjoined elements. Their evaluation then forms the basis for preference scores assigned to these distinct components which we assume would have resulted if the respondent were able to assess them correctly.

Put quite simply, instead of asking them to compare apples with pears, we change them all into oranges making the evaluation much easier – the fruit machine analogy is a good one in that you are aiming to change differing elements into the same easily recognisable structure. The jackpot of course is the understanding gained in terms of a most preferred unique basket of product attributes and/or the most acceptable pricing point.

The technique allows researchers to rank and quantify a myriad of brand elements, in terms of pricing, benefits and new product development. It enables the capacity to deliver at many different junctions during the product cycle from market entry analyses, optimal pricing points and competitor pricing.

But that’s not all conjoint allows us to do. The tool unlocks the soft emotional characteristics of the human condition and so what appears at first sight to be a rigorous statistical engine is actually taking the role of empathetic listener, sympathetically unearthing sentiments that respondents are otherwise unable to express.

So what if we applied the technique to the more qualitative brand issues? The more sensitive nuances like equity, protectiveness, perceived quality, stickiness and brand associations can all feature in the conjoint model and provide really insightful outcomes.

And what if then we applied this to a brand tracker? This is where it really does get powerful: imagine being able to follow the variances of a brand continuously, every campaign, every news story. Ultimately, every subtle nuance in linear time is able to be analysed.

Powerful indeed, but the process of designing the integral building blocks, scripting the attribute bundles within an online survey platform and then being able to analyse the output accurately can seem very daunting. However, at Nepa we argue that this does not need be the case – with our end-to-end provision you can take the guesswork out of the technique and demystify the journey.

So essentially conjoint focuses on the trade-off people make between different attributes and price levels if required. These trade-offs are often made implicitly without thinking and can be based sometimes on gut feeling, sometimes after very rational and thought through consideration and at other times it’s simple heuristics that guide people through the decision processes.

The bonus with conjoint is that we don’t have to care too much about this. Since conjoint analysis is based around a concept where we put people in a series of choice tasks and see how they respond to different contexts we do not need to rely on the introspective capability of consumers which many other methods do. We simply look at the results and derive the underlying choice process backwards. This is similar to what researchers do when they use other multivariate techniques such as regression, structural equation modelling, and PLS (Partial Least Squares regression – a form of analysis that finds a linear regression model within the data).

However, conjoint has yet another advantage over these methods: their purpose is to give good predictions on general, universal, stable, long-term trends. They rely on big numbers and every respondent is just one “case” in the analysis, we use these cases to draw conclusions from averages and correlations. This is reasonably OK if you are looking to predict the behaviour “on average” but will be sadly lacking when trying to guess how individuals will actually behave.

Conjoint instead, if we allow for some simplification, does a multivariate regression on each individual. Since every respondent is subjected to a series of choice tasks, and these choice tasks constitute the “cases” in the multivariate analysis, we are able to produce one individual choice model for each respondent. These individual models are then used to predict the behaviour of groups of people or the behaviour of the market as a whole.

Conjoint analysis is actually a whole family of similar techniques which are conceptually linked together by their experiment-like nature. At the core of all of these techniques is that the respondents are put through a carefully designed experiment which has the combined advantages of being easy and relatively entertaining to participate in and at the same time produces very rich data which can be used in an array of ways to simulate, predict and understand the behaviour of individuals, segments and markets alike.

The fact that the industry has become used to accessing online as a way of data gathering sits perfectly within the conjoint process. Complex matrices can be built directly into the platform meaning that the respondent only sees the product offering and decision tree appropriate to them.

Nepa has conducted a multitude of conjoint studies for clients ranging from market research agencies to brand owners, and this experience and the linear execution of our processes has led us to believe that the technique could be used much more regularly and applied to many more research objectives than is presently carried out.

Making the tool accessible to all is the first step in raising its profile and along with other less utilised forms of multivariate analysis we hope to see a lot of the associated myths dispelled and that black box finally opened for all to benefit. Our user-friendly desktop dashboard – branded N-Visualize – takes the strain out of analysis pointing you in the direction of the answers you are looking for and allowing intuitive display of complex data for easy and quick comprehension.

A long overdue renaissance for such a tried, tested but perhaps underrated research tool can only enhance the research industry – and provide clients with the all too elusive actionable insights they crave!