top of page

Conjoint Analysis Autonomous Vehicles

Our client, a prominent competitor in the automotive industry set to become an EV provider is interested in learning which attributes of a self-driving car are most influential in shaping customers’ preferences. Through conjoint analysis, we discerned the importance of each attribute in shaping customer preferences and explored potential market segmentation based on these preferences. Find out where they should focus their efforts!

Background & Problem Definition

In the rapidly growing market of autonomous vehicles, understanding customer preferences towards various product features is crucial. Our client, a prominent competitor in the automotive industry, is interested in learning which attributes of a self-driving car are most influential in shaping customers’ preferences. The features under consideration span several key aspects of autonomous vehicles: safety, level of autonomy, connectivity, energy efficiency, and comfort. These elements range from advanced and standard safety measures to varying degrees of autonomous operation, types of connectivity, and energy efficiency levels, as well as a spectrum of comfort and luxury options. While the price is not an attribute for study, the client intends to competitively position their product within the price range of $55,000 to $75,000.

To investigate this, we analyzed survey data where 52 respondents ranked 12 distinct car profiles, each with a unique combination of these attributes. Through conjoint analysis, we discerned the relative importance of each attribute in shaping customer preferences, estimated the part-worth utilities, and explored potential market segmentation based on these preferences. The client is also interested in developing two new products with specific attribute combinations. We evaluate the projected market share of these products against existing competitors, namely a Tesla model and a Hyundai Ioniq with specific features. The goal of this analysis is to provide strategic recommendations on product development and pricing based on the findings, thereby strengthening our client’s position in the autonomous vehicle market.

Methods & Diagnostics

Preprocessing & Feature Selection

The attributes of the original data frame were transposed into separate data frames in order to make modeling approaches simpler. A quick check of these data was performed to seek out missing values or other nuances that would impact analysis. Missing values or other impediments were not detected. 

1. Most Important Customer Preference Attribute: 

In performing the conjoint analysis, our results showed five attributes in decreasing order of importance (see Figure 1 in the Appendix).

The most important attribute that drives customer preference in this analysis is “Comfort”. We found this notable because although we initially expected “Safety” and “Autonomy” to be major drivers due to the nature of self-driving cars and the reliance on the technology, it seems that customers value “Comfort” the most. It is possible customers are assuming a base level of safety and autonomy for these products and that these aspects are less influential. Current events around self-driving car safety flaws and development processes also led us to believe safety would be a top priority for customer preference.

In contrast, “Comfort” being most important would offer more tangible and immediate benefits that not only customers perceive to be beneficial but are also readily more achievable as a product strategy. Furthermore, “Connectivity” came as the second most important attribute which indicates to us that customers value the ability of the vehicle to be connected to other devices or systems. Which gives a hint that the interest might rest on associated features such as internet connectivity, smart device integration, etc.

“Energy” ranked as the third most important attribute. This tells us customers place a significant value on the energy efficiency of the car and/or the type of energy it uses (electric, hybrid, etc.). While the importance of “Safety” and “Autonomy” were lower than we expected, they still contributed significantly to overall preference. Overall, the car product should focus on enhancing comfort, connectivity, and energy while also taking into account safety and autonomy.

2. Part-worth Utilities Insight: 

The part-worth utilities are also known as the attribute utilities, which are numerical coefficients determined by the conjoint analysis. The coefficients measure the relative importance of each level for each attribute, essentially quantifying the contribution of each attribute level to the overall product or service that is being preferred. In this analysis, the part-worth utilities are the coefficients derived from a regression model. They explain the degree to which each level of each attribute influences the overall preference. The estimated part-worth utilities can be found in Figure 2 of the Appendix.

Insights Drawn:

  • Safety: The data suggests a preference for Advanced Safety as compared to Standard Safety. Customers highly value safety, particularly more advanced measures.

  • Autonomy: The relatively low and comparable utilities for Medium Autonomy and High Autonomy imply that the degree of autonomy may not actually be as important of a factor in influencing the preferences among customers.

  • Connectivity: The significantly higher utility of Advanced Connectivity over Standard Connectivity suggests customers place a premium on connectivity capabilities.

  • Energy: Low Energy has a higher utility compared to Moderate and High Energy. This points towards a preference for cars that either have better energy efficiency.

  • Comfort: With the highest positive utility among all the levels, Basic Comfort emerges as the most influential factor. This aligns with the previously derived attribute importance findings, which identified Comfort as the most critical factor. A noteworthy point, however, is that Premium Comfort demonstrates a negative utility, which suggests customers might not perceive the potential added cost associated with additional comfort features as valuable. 

3. Natural Segments Analysis:

Our segmentation analysis identified three distinct clusters, which shows that there are natural segments of customers for self-driving cars based on the attributes analyzed (see Figure 3 in the Appendix).

  • Cluster 1: This segment consists of 22 individuals. The cluster mean shows the preferences within this segment are safety, autonomy, and energy efficiency. We can create a customer profile for this group and refer to them as “Safety and Autonomy Seekers” – individuals in this segment prioritize advanced safety features and higher autonomy in their self-driving cars coupled with energy efficiency. 

  • Cluster 2: This segment is composed of 20 individuals. The cluster mean for this group suggests strong preferences for attributes related to energy and connectivity. Our customer profile would be the “Connectivity and Energy Enthusiasts” group – they value advanced connectivity features and high energy efficiency in their self-driving vehicles.

  • Cluster 3: This is our smallest segment of 10 individuals which shows a clear preference for attributes pertaining to comfort and autonomy. Our customer profile should be the “Comfort Lovers” – these customers place a high value on comfort and higher autonomy in their product choice.

Overall, these patterns suggest that our client could benefit from considering differentiated pricing strategies for these segments. For example, premium pricing can be considered for advanced safety features for Cluster 1, while the focus is on advanced connectivity features for Cluster 2, and for higher add ons of comfort for Cluster 3. The client should consider the competitive landscape, price sensitivity, operational complexity, and costs to attract a larger customer base. 

4. Product Analysis:

The market share simulation results from the conjoint analysis for the four different products – the two proposed, the current Tesla model, and the Hyundai Ioniq – were calculated based on the Bradley-Terry-Luce (BTL) model and the Logit model, which are commonly used methods for predicting choice probabilities.  The total utility and maximum utility scores provide an inference of the aggregate desirability and best possible desirability of each product offering, respectively (see Figure 4 of the Appendix).

  1. First Product (Standard Safety, High Autonomy, Advanced Connectivity, Moderate Energy Efficiency, Premium Comfort): This product has a total utility score of 6.59 and maximum utility score of 21.15, leading to a predicted market share of 28.62% based on the BTL model and 21.86% according to the Logit model. This product, with its emphasis on high autonomy, advanced connectivity, and premium comfort, appeals to a significant portion of the market.

  2. Second Product (Standard Safety, Medium Autonomy, Standard Connectivity, High Energy Efficiency, Basic Comfort): The second product shows a total utility score of 5.59 and a maximum utility score of 13.46. It achieves a predicted market share of 19.09% according to the BTL model and 17.34% from the Logit model. This product, with high energy efficiency as its standout feature, appeals to a smaller but still substantial market segment.

  3. Tesla (Advanced Safety, Medium Autonomy, Advanced Connectivity, Low Energy Efficiency, Standard Comfort): Tesla's current offering demonstrates a high total utility score of 9.99 and a very high maximum utility score of 59.62. This translates into a predicted market share of 38.56% as per the BTL model and an impressive 58.49% according to the Logit model. Despite the low energy efficiency, the advanced safety and connectivity features appear to be driving preference for this product.

  4. Hyundai Ioniq (Advanced Safety, Medium Autonomy, Standard Connectivity, Moderate Energy Efficiency, Premium Comfort): The Hyundai Ioniq, with a total utility score of 4.23 and a maximum utility score of 5.77, has predicted market shares of 13.74% and 2.30% according to the BTL and Logit models, respectively. This offering seems to hold the least appeal among the four options, likely due to standard connectivity features.

We can conclude that the market share simulations suggest a heterogeneous market with varying preferences for different attributes. Both proposed products would command considerable market share, but Tesla's current model remains the market leader. We recommend the company conduct further research to bring more market data and/or attribute levels.

5. Recommendations:

Based on our findings from this conjoint analysis, we recommend the following strategy:

  • Product Configuration: Our analysis underscores the importance of balancing a mix of attributes to maximize customer appeal. Safety and comfort emerged as critical elements for customers, emphasizing the necessity for manufacturers to focus on these elements in their designs. Additionally, advancements in connectivity also drive product preference, indicating a need to invest in this area. 

  • Customer Segmentation: The results of the segmentation analysis suggest the existence of distinct customer clusters with varying attribute preferences. Manufacturers should use this insight to tailor product offerings to specific market segments. For instance, one segment may prioritize high autonomy and premium comfort, while another may value energy efficiency and basic comfort. By offering product variants targeting these segments, manufacturers can maximize market penetration and customer satisfaction.

  • Pricing Strategy: The segmentation results could also inform pricing strategies. Considering the varying attribute preferences across segments, manufacturers might consider value-based pricing. For instance, a product with high autonomy and premium comfort could be priced higher for the segment that values these attributes.

  • Product Launch Strategy: The market share simulations indicate potential competitive scenarios for new products. Manufacturers should use these insights to develop launch strategies that position their products effectively against the competition. For instance, our findings suggest that a new product offering with high autonomy, advanced connectivity, and premium comfort could significantly compete against Tesla's current model, which holds the largest market share.

  • Research & Development: Research and development should always be weaved through the heart of any manufacturer or company for that matter. Because while the results provided valuable insights, the discrepancies observed between the two market share prediction models underline the need for further research and development. This could involve the integration of additional market data, refinement of attribute levels, or application of different conjoint analysis models.

We believe this conjoint analysis provided crucial insights for the manufacturer's strategic decision-making, from product development to marketing and pricing strategies and that by understanding and responding to customer preferences, the client can design and market products that meet customer needs.


Figure 1: Attribute rankings with percentages

Figure 2: Part-worth utilities of attribute levels

Figure 3: Visualized cluster segmentation results

Figure 4: Market share simulation results table

Project Gallery

bottom of page