W5 on Conjoint - Part 2: How It Works

W5 on Conjoint is a three-part series exclusively available on The W5 Narrative. Click here to read Part 1: Conjoint 101.

Where were we?

In Part 1, W5 introduced Conjoint and explained how it helps brands produces superior products, impactful branding, and sound pricing strategies by quantifying individual product attributes to generate a uniquely reliable understanding of consumer preferences.

In Part 2, we’ll take a closer look at how Conjoint research is conducted and share our tips for ensuring a successful study design.

Remember, Conjoint strives to mirror real-world choices consumers consider when making a purchase by presenting numerous factors and alternatives aimed to capture how respondents would think and act in a real-life purchasing scenario. While the concept of Conjoint is straightforward, how you get there is not.

Okay, let’s get into the nitty gritty of how this works.

To design a successful and insightful conjoint analysis, there are four key steps in the process to be considered and executed correctly.

Let’s look at each of these steps in detail…

1.       Consider the Market and Organize Your Input

Through discussion, review of relevant previous research, and close collaboration with our clients, W5 excels at helping clients “get everything out on the table” ahead of designing a conjoint exercise.

In addition to selecting the product category and the range of options to cover, it is necessary to identify each attribute to assess ahead of time – and what the levels (variant options) of those attributes are.

At this point in the process, we develop a Microsoft Excel spreadsheet of columns for the attributes and rows for the levels enabling us to visualize how much and what we want to cover in the Conjoint.

Let’s go back to the bakery selling doughnuts example we presented in Part 1. An example of the Excel document might look something like this:

2.       Determine the Right Conjoint Approach

Once the spreadsheet of product attributes and levels is developed, it is important to take a step back to consider any possible combinations that do not – or cannot – work in the real world. There may be certain options that do not make sense at certain price points – or certain features that cannot exist in combination with others. This is where restrictions come in.

You can set restrictions limiting combinations that will be presented. W5’s expertise will help guide the best approach when consider restrictions (i.e., presented in the model but then avoided and controlled for analysis). In the end, the goal is for the conjoint results to make sense in the broad market – not just the survey environment, so these considerations of actionability are very important.

Moving into the next stage of the process, W5 chooses the type of conjoint design appropriate for the model. This generally depends on the number of attributes to be assessed, the nuance between the levels to be presented, sample size, and estimations of model reliability.

Most W5 conjoint designs are either choice-based conjoint (CBC), adaptive conjoint analysis (ACA), or a hybrid of these two types: adaptive choice based conjoint (ACBC).

Another critical design choice is whether to present full profile or partial profile conjoint questions in the model. This is a choice whether to show all or just part of the interchangeable options for the product screen-to-screen. This decision too, is made based upon the breadth and complexity of the input to be assessed in the model – especially the number of attributes and expanse of levels per attribute.

3.       Field the Survey

Now that the heavy lifting of design is done, it’s time consider the conjoint as a survey tool. Remember, the conjoint exercise (sometimes called “experiment”) is the set of questions within a survey that present various sets of product combinations along with a preference question to capture all consumer interest and demand data.

While the questions are somewhat redundant, sometimes presenting difficult and limited options to choose, this is intentional. It is important to gauge consumer sensitivity and receptivity to a range of alternatives, not just a reaction to a packaged whole.

In addition to the conjoint preference question, W5 often also directly asks a yes/no question: “Is this really an option you would purchase?”  The statistics from this question may or may not be modelled into the conjoint analysis (depending on the situation and response levels) but do provide respondent comfort and valuable insight regardless.

It’s important to remember we’re seeking to reliably model a real-world trade-off and purchase decision. This means it is critical to ensure that all survey records retained for analysis represent real, qualified, engaged, and attentive respondents. As with all W5 surveys, we monitor, flag, and replace lower quality and less engaged respondents, ensuring the analysis is setup to succeed based on thoughtful and deliberate response.

4.       Conjoint Analysis and the Market Simulator

Once data has been collected, cleaned, and compiled, W5 prepares for conjoint analysis. Conjoint analysis is fundamentally a model based upon probability, rather than descriptive statistics. Instead of seeking a percentage of the market who say they prefer an option (e.g., 65%) we aim to model consumer preference, prioritization, purchase likelihood, and price sensitivity based on data reflecting the interaction of all the variables we assessed in the conjoint model (the attributes and levels).

For conjoint analysis, W5 leverages Hierarchical Bayesian statistics, linking the survey data to iterative probability-modelling procedures. The conjoint data is processed 20,000 times via Sawtooth Software, ultimately yielding clear, reliable scores called “part-worth utilities.” These are akin to “utils” many of us learned about in college economics classes.

W5 analyzes these scores and outcomes to produce a strategically focused and intuitive report and presentation delivering insights, takeaways, and “so-what?” implications. W5 also leverages an interactive Market Simulator during reporting, a Sawtooth Software tool where you can play “what-if” scenarios and answer the questions from your conjoint study.

Phew, that was a lot.

Yes, it was. But don’t worry. W5 has over twenty years’ experience executing Conjoint research and will be a strategic partner (and coach) to get you from organize to analysis.

We know your brain is on overload so we’ll call it a day. Be sure to come back for Part 3 where we’ll highlight one of the most fun (and surprisingly complicated) conjoint studies W5 has conducted. And yes, doughnuts are involved.

W5 on Conjoint is a three-part article exclusively available on our blog, The W5 Narrative.