As a data scientist, there is one sentence that I utter probably more often than any other sentence.
First, let me set the scene for you.
A client (or potential client) has identified a need or a gap in their knowledge, and they’re coming to us Sprocketeers to get the party started on a new project. So many times, the conversation starts like this:
Client: “April, I need an [insert name of statistical method here].”
While I love that my client has thought about methodology, I find that beginning the conversation by anchoring on a particular statistical method immediately boxes us in, and we run the risk of falling short or missing the opportunity to deliver insight or analysis that truly answers a client’s real question or meets their true need.
Here’s where my sentence comes in. You ready for it?
Me: “That’s great! But let’s take a step back. Just for a moment, instead of telling me what method you want, let’s talk about what you want to do with it.”
Bam. There it is.
Instead of anchoring on a particular method, I invite my client to describe to me what their goals are. What are they trying to accomplish? Essentially, I invite my client to trust me to help them identify the best method (or more often, methods) that will get them the answers, insight, and direction they need to move confidently forward.
There are few methods that anchor or box clients in more than the conjoint analysis.
Conjoint Analysis: The Sparkly Unicorn
I don’t know what it is about the conjoint analysis, but it has somehow earned the title of the mystical unicorn of research and statistical methodology, and when clients encounter this sparkly creature, they have an extremely hard time seeing beyond it.
Don’t get me wrong, I’m not knocking conjoint. It’s a powerful, useful tool. However, to be clear, I find that it is rarely the end-all of what a client is hoping to learn. More often, it is one important component in a project that will truly help a client meet their goals.
Curious? Read on. Let’s start with the basics.
What is Conjoint Analysis?
Conjoint analysis (sometimes called discrete choice analysis), is a statistical technique that is used to uncover the degree to which customers value a set of defined product or service characteristics and how they contribute to customers’ overall decision to purchase.
Respondents in a conjoint study are shown different ‘cards’ containing configurations of sets of product features and pricing. There are a number of different variations, but in choice-based conjoint, they indicate which ‘card’ they would choose out of the set shown to them. Across multiple trials and multiple respondents, you can determine which product features and prices are more favorable to potential consumers. You can even determine the statistically optimal product configuration!
That’s some powerful stuff!
…..but we can do better. We just need to tear our eyes off of the sparkly unicorn and open our minds to other methods that can help round out the insights that a conjoint analysis can uncover.
Beyond the Sparkle
Although there are many, I want to highlight just two additional questions that some modification and addition to a traditional conjoint analysis can help answer.
1. Even if I use the results of the conjoint analysis to put together the optimal product configuration and pricing, how do I stack up against my competitors in the market?
A traditional conjoint analysis focuses on a given business’ product. That is incredibly powerful insight, but it doesn’t tell you where and to what degree you are vulnerable against your competitors’ products.
There are ways to augment a conjoint analysis (‘hack,’ if you will) to push the insight you get even further. Wouldn’t it be valuable to know that a competitor’s feature stack that isn’t even on your radar drives consumers to purchase, over and above the features your product contains? It might even change the future trajectory of your product development.
A traditional conjoint analysis would leave you in the dark.
2. The conjoint analysis showed me what my optimal product configuration is, but will that optimal configuration perform well IRL (in real life)?
At Sprocket CX, we combine the intelligence of data science with the elegance of design thinking. We like to say that ‘you can’t analyze your way into the future.’ As such, a test and learn approach to determining whether your optimal product configuration will actually perform well on real, live consumers who are in-market for your product can be extremely valuable.
We believe in constructing small, targeted experiments to help iterate toward a live solution that we’re confident will perform well in the market. Again, a traditional conjoint analysis wouldn’t get you there.
What’s Your Goal?
In this post, I’ve barely scratched the surface. Nothing gets us more jazzed than to dream up the optimal set of methods to help our clients achieve their unique goals. What’s your goal?
Give us a shout, and we’ll help you dream up the right mixture of methods (with the right degree of sparkle) to help you meet your goals and move forward confidently.