Just under 10% of the 250-strong team at Dig Insights is exclusively dedicated to advanced analytics. This team is dedicated to developing new and improved ways of generating and using data. This includes embedding AI into our workflows, developing a patented market simulator that automatically translates Upsiide data into forecasts of share of choice, source of volume, incrementality and cannibalization and reinterpreting classic research models such as TURF and conjoint. Thanks to our Advanced Analytics team, we have tools like Upsiide’s Idea Map and Market Simulator, Virtual Market, Virtual Menu, Share Optimization (ShOp), and numerous custom data analytics and visualization tools.

What is conjoint and how can it be improved?

Conjoint is a discrete choice methodology used in market research since the 1970s. It is used to uncover how consumers make trade-offs between competing product attributes. With conjoint, consumers are shown in-market products and products that could enter the market. They are asked which they would choose. Based on those choices, conjoint derives what drives consumer choice. Conjoint has evolved over the decades, fueled by advances in technology and statistical techniques. With survey completion on mobile devices rising, there is a need for conjoint to evolve further and create a mobile-first respondent experience.

The experience on laptops involves respondents completing two steps:

1. Compare multiple options and choosing their preferred one.
2. Indicating if they would buy the option that they chose.

The issue we’re solving is how conventional conjoint is adapted for mobile devices. To complete step 1 on a phone, respondents must scroll horizontally to see the competing options. This forces respondents to make a product choice based on recall of product details that are not visible on the screen.

Respondents then indicate if they would buy the option that they chose just as they would on a laptop. Improving the respondent experience on mobile devices will ensure that conjoint remains a viable methodology.

Upsiide conjoint

Dig Insights has developed the capability to conduct conjoint on Upsiide, our proprietary platform dedicated to innovation optimization and testing. Upsiide is a mobile-first, gamified approach to capturing behavior, emotions, and thoughts. It leverages Dig Insights’ expertise in understanding and predicting consumer decisions.

With Upsiide, we split the two steps from the traditional approach into two separate exercises that reflect how consumers make decisions in the real world.

Interest (Exercise 1)

In the real world: consumers encounter an innovation. In an instant, they lean in or move on.

In Upsiide: we ask people if a product or service is relevant to them (swipe right) or not (swipe left). They see one product at a time vs. three or more in a conventional conjoint.

This radically simple interface encourages intuitive reactions and identifies innovations with the potential to break through in a crowded market.

Commitment (Exercise 2)

In the real world: consumers choose an innovation or a competing alternative.

In Upsiide: we present people with pairs of products or services that are relevant to them personally. They choose a favorite.

Commitment identifies products that have the potential to shift choice in a competitive context.

Comparing the results of Upsiide conjoint vs. traditional conjoint

We completed three parallel tests to compare Upsiide conjoint vs. traditional conjoint.

There are two key metrics:

1. Model similarity – the goal is to deliver comparable results when conjoint is conducted on Upsiide vs. conventional conjoint
2. Respondent experience – the goal is to deliver a superior respondent experience

Model similarity

We observed a high correlation of results in each of the three parallel tests. The charts below show the correlation between utility scores in the Upsiide and traditional conjoint approaches. Utility scores reflect the ranked performance of the ideas evaluated with each attribute (e.g., within the Brand attribute, we evaluate the individual brands DiGiorno, Red Baron, etc.). Utilities rank those brands from strongest to weakest.

Our first parallel test was conducted in Nov. 2022 and focused on the coffee category. We surveyed 800 US coffee drinkers and split evenly between the two approaches. The test included four attributes (Brand, Origin, Roast, and Price), with four levels within each attribute.

Our second parallel test was conducted in Aug. 2023 and focused on restaurant pizza. We surveyed a total of 1,400 US consumers who regularly order pizza from a restaurant. The sample was split evenly between the two approaches. The test included five attributes (Restaurant, Type, Crust, Price, and Yeast), with five levels within each attribute.

Our third parallel test was conducted in May and June of 2024 and focused on frozen pizza. We surveyed a total of 1,200 US consumers who regularly buy frozen pizza. The sample was split evenly between the two approaches. The test included five attributes (Brand, Type, Crust, Claim, and Price). There were 6-8 levels within each attribute.

Utility scores (shown above) report the performance of levels (e.g. all tested brands, all tested claims). Importance scores reflect the importance of attributes (e.g., the importance of brands or claims overall to determining choice).

In this example, we see that importance is remarkably similar across Upsiide conjoint and conventional conjoint.

Respondent experience

We evaluated the respondent experience in the first two parallel tests using a combination of attitudinal and observed measures. The number of conjoint tasks completed is provided for context.

Enjoyability – Rating the study from 0 (boring) to 100 (enjoyable).

Reasonable – Rating the study from 0 (overwhelming) to 100 (reasonable).

Time to complete (median) conjoint tasks.

The results for both parallel tests demonstrate that the respondent experience is similar for traditional and Upsiide conjoint. The most compelling difference is the time to complete. Across the two parallel tests, Upsiide conjoint took about 10% less time than the traditional conjoint.

Note that Upsiide conjoint involves more tasks (12 vs. 52 in the above examples). But each conjoint task is vastly simpler and more intuitive, involving swiping left (dislike) or right (like) on a single idea and then trading off between pairs of liked ideas. This compares to choosing between grids of three or more product choices in a conventional conjoint.

The latest research (behind a paywall, sorry) from Sawtooth indicates that the form of conjoint used in Upsiide can be reliably conducted with 20% fewer tasks. We evaluated this in our third parallel test and observed a 32% reduction in time to complete while maintaining a 0.97 correlation. This implies that our mobile-first design has effectively simplified the respondent experience while maintaining data quality.

Conclusion

We have created a highly effective mobile-first approach to conjoint by moving from a grid interface to two intuitive exercises. Upsiide conjoint:

• Takes less time to complete
• Creates a mobile-first experience, opening conjoint studies to a broader audience
• Produces results highly correlated with traditional conjoint
• Is rated similarly to traditional conjoint on attitudinal measures