Pushing the Boundaries of Conjoint
Listen to the Audio Version here
When it comes to product and price optimization, there is an unofficial gold standard that experts refer to: Conjoint. The method is most often used when developing and launching new products, or when re-launching a product or family of products.
To inform product development, pricing, and market communications, companies need to know how well certain features and price points drive or inhibit purchase decisions.
"These examples show how impractical Conjoint can be“
Example 1 - Apple Vision Pro:
A year before its launch, we at SUPRA put the Apple Vision Pro through the Implicit Intelligence research process because it pushed the boundaries of traditional research approaches. First of all, this Apple product opened up a new category and there was only one other product on the market that could be used as a reference point (Metas VR glasses) - how do you set up a conjoint with that? To make matters worse, the number of essential features to be evaluated was beyond the scope of any reasonable conjoint approach.
The feature set we tested was:
Spatial user interface, intuitive control by eyes, gestures and voice, integration of physical environment with user interface, user's eyes are visible to others, ultra-high resolution display, infinite workspace, 3D photo and video capture, spatial video telephony, high-quality aluminum and glass materials case, 3D video, games and live events, App Store has many 3D games and movies, 2 hours long battery life.
There was also concern about how a simple bulleted list of features (the way Conjoint displays a product) could convey well enough what the product is really about.
Example 2 - footwear:
An international footwear brand wanted to reposition and re-price its six key products and understand the role of key attributes.
Here is the list of 30 features important to the marketing team:
Traction, Suitable for the outdoors, Arch support, Lightweight, Popular brand, Comfortable, Craftsmanship, Water-resistant, Waterproof, Attractive design, Anti-fatigue, Easy to mix and match, Cushioned, Suitable for everyday, Suitable for a variety of occasions, Padded collar, Hand-sewn construction, Premium leather, Fits well, Warm, Premium construction, Iconic design, Suitable for hiking, Rubber Lug Outsole, Breathable, Insulated, Sustainable materials, Durable
Not only was the number of features a challenge for the conjoint, but the assumption of competitive products seemed unrealistic to the team. The products are mostly sold in branded stores where there is only internal competition, while in Internet stores the competition is almost unlimited. At the very least, there is no clear, well-defined set of competitors that would realistically model a typical purchase decision.
Example 3 - credit cards:
A bank planned to change the credit cards it offered to account plans. This would change not only the features, but also the monthly card fee. It was not clear how to promote the change or how to price it. Since the prospects were their own customers, it was not easy to structure a competitive conjoint task.
In addition, the characteristics of interest were well beyond the scope of conjoint:
Accepted worldwide, Apple Pay supported, eCommerce payments, full spending control, 2-factor authentication, PIN of choice, amount reservation (e.g. in hotels), cash at HVB ATMs, fraud prevention, self-service functionality, change PIN online, Google Pay supported, virtual credit cards, app push alert instead of SMS, order blocking & limits online, sending money to friends, delayed account debit, immediate account debit
Limitations of Conjoint
The advent of conjoint measurement introduces a major advantage. The approach structures a choice task in a way that makes it difficult for respondents to solve the task rationally. It forces them to use part of their gut or intuition, as in real life.
This advantage certainly disappears if the number of features is too small. If you look at typical conjoint studies, you will see that they often use only two or three features. Too often, only price and brand are used.
The original conjoint measurement from the 70's is hardly used anymore. Mostly choice based conjoint (discrete choice modeling) is what research agencies offer. Here, too, we have a wide range of applications. With or without non-option? Sum or Choice Based Conjoint? Limit or HB Conjoint? Conjoint requires a lot of experience, both to set it up and to analyze it properly.
The limitations of the approach have already become clear in the examples above:
It allows for a maximum of 7 attributes, while many products require much more.
It provides a rudimentary product description in the form of bullet points and therefore does not provide a realistic representation of real products. This alone affects the validity and predictability of the results.
The measured price demand elasticity is strongly biased by design choices, which is why coinjoint sometimes underperforms even the simplest pricing techniques such as Gabor Granger. Price levels are also biased as price anchors.
Is MaxDiff a cure?
MaxDiff is a survey approach that asks for preferences for all pairs of features in the feature set. Based on this, it computes an importance ranking. This approach is typically recommended when the number of features exceeds the capabilities of conjoint.
While it is a practical and useful method that complements conjoint well, it has serious limitations of its own:
It does not measure a price demand curve. It measures only relative importance and cannot determine whether a feature has an absolute negative or positive value.
It's an explicit measure of importance - as such, it overvalues hygienic features and undervalues pleasurable features. As such, it is a biased measure of importance.
"Neither Conjoint nor MaxDiff meet the needs of many product teams"
The Rise of Implicit Intelligence
95% of a purchase decision (B2C and B2B) is driven by the unconscious, implicit associations that occur within milliseconds of a product presentation. While it takes seconds to form a rational opinion, our immediate gut instinct guides how we rationally process information and what information we focus on. This also applies to price perception itself. The "gut" tells us if a price is too high or too low, regardless of what price knowledge we actually have. Unleashing this implicit association is the key to reliably estimating the price demand function as well as the perception of product features.
This is what Implicit Intelligence does. The name comes from the fact that the approach combines implicit research (a form of the Implicit Association Test - IAT) with artificial intelligence (specifically Causal AI).
This AI is used in two ways: First, it eliminates survey bias introduced by the price anchor effects of the highest and lowest price points tested. Second, the AI serves as a modern driver analysis, providing insight into the leverage that improving the perception of a feature would have.
This is the exact procedure of the method
Introduction: The respondent can view a detailed product description with an image and perhaps a video. The intent is to provide as much information as the customer would optionally consume at the point of sale.
Price IAT: The product name with picture is shown next to one of seven price points. The respondent is then asked whether the price fits a particular attribute, such as "fair," "reasonable," "I buy.
Context: Customers who love the brand or have already bought the product are more likely to buy. We control for the biggest external drivers of purchase so we can recalibrate.
Feature Importance IAT: In an IAT setting, we ask whether the feature is more or less important. This is the implicit way to measure importance, with the advantage over MaxDiff that it measures absolute, not relative, importance. This analysis will reveal the "importance of having the feature".
Feature Performance IAT: In an IAT setting, we ask whether the respondent perceives that the product has a particular characteristic. This is the key input for the causal driver analysis. This analysis will reveal that the "lever of improvement" may be fundamentally different from the importance of having the feature.
Brand Architype IAT (optional): In an IAT, we ask respondents if they associate the 12 archetype-related attributes with the brand. This is another input to the causal driver analysis. While measuring attributes can inform WHAT to write in the communication, the brand archetype largely informs HOW to write and how to show it.
Depending on the number of characteristics, the questionnaire typically takes 7 to 10 minutes to complete. It is therefore no more demanding for respondents than conjoint questionnaires.
The implicit price acceptance test is actually an accurate measure of willingness to buy and willingness to pay (at a given price point). Causal Driver Analysis can now use this information as the ultimate target variable. It even provides seven data points and facets of purchase intent, making the whole exercise statistically very robust.
By bolding different assessment exercises, such as Feature Assessment, Brand Architype Assessment, or even a Touchpoint Contact Assessment, we are able to understand the impact of features, architype, or touchpoints on the purchase decision. All in a single exercise.
"Implicit Intelligence Combines Implicit Measurement with Causal AI Driver Analysis“
Result dashboards at Supra.Tools
The price demand function is accessible in a dashboard. Sliders allow the user to make assumptions about the cost of goods sold, sales tax, and retail margin. With these inputs, the dashboard shows how the brand's profit depends on pricing. This allows brands to make profit-maximizing pricing decisions.
The Focus View dashboard contrasts the "importance of having it" (derived from Implicit MaxDiff) with the "lever to improve it" (result of Causal AI Driver Analysis). The example here shows the various features of the Apple Vision Pro. Exciting insights here are that features like "ultra-high screen resolution" and "2h battery life" are very important, but there is no lever to improve them. It shows that there is no need to improve them even though they are "very important".
The Improve View dashboard shows the 12 features (blue) and the eight common barriers to innovation adoption (green) in a matrix diagram that contrasts the "lever of improvement" with the actual perception (performance) of the features. The analysis showed that the features "3D video, games, and live events" and "intuitive control with eyes, gestures, and voice" have a huge lever, but in reality, prospects have doubts about the product's ability to deliver. These are the core areas that Apple needed to address in its communications efforts to make the launch a success.
What the platform does not provide is a market simulation - a feature that many conjoint software solutions offer. Such a simulation requires the evaluation of a competitive set of products.
If brands choose to run competitive products through the implicit intelligence testing process, we at Success Drivers will run a market simulation outside of this tool. If needed on a regular basis, we can implement this feature in Supra.Tools.
This is when Implicit Intelligence beats Conjoint
In the evolving landscape of market research, traditional conjoint analysis has long been the go-to method for product and price optimization. However, we present a disruptive approach, Implicit Intelligence, that promises to surpass the capabilities of conjoint analysis. The article discusses the limitations of conjoint analysis, such as its inability to handle large numbers of product attributes, provide realistic product descriptions, and accurately measure price elasticity. These limitations often lead to biased results and reduced predictive power.
Implicit Intelligence provides a solution by leveraging the subconscious, rapid associations consumers make with products. It integrates Implicit Association Testing (IAT) with Artificial Intelligence, specifically Causal AI, to overcome the biases and limitations of traditional survey methods. By capturing instinctive responses to product features and prices, it provides a more accurate reflection of consumer behavior and preferences.
The Implicit Intelligence process involves multiple steps, including detailed product introductions, assessing the importance of price and features through IAT, and controlling for external purchase drivers. This approach not only measures the absolute importance of features (unlike MaxDiff, which measures relative importance), but also allows for causal driver analysis to understand the impact of improving feature perceptions.
I also highlighted the practical applications of Implicit Intelligence with insights from testing products such as the Apple Vision Pro. The methodology provides detailed dashboards for analyzing price demand functions, the importance and potential for feature improvement, and identifies areas for communication efforts to improve product launch success.
Brands using it with great success
Innovative insight leaders have been using it for several years and share their positive experiences:
Bodo Drieschner of Hypovereinsbank/Unicredit "The Supra Price & Product Optimizer helped us analyze our questions about account packages and identify an optimal setup. Particularly impressive is the profit simulator, which makes it easy to understand the impact of costs on the optimal price.
Anne-Laurence Picard, Sr. Director, Consumer & Shopper Insights, EMEA TIMBERLAND: "The Supra Price and Product Optimizer approach has helped us sharpen the role and positioning of our footwear icons in the portfolio and their communication at the point of sale by identifying their respective consumer value drivers. The Price Optimizer tool has also helped us unlock pricing opportunities and effectively use price as another positioning differentiator based on relevant and meaningful consumer benefits/features."
These innovative insight leaders operate by a simple credo: If it makes sense and has great potential, just try it.
This is how you 10x your Insights.