Personalization of Experience in Action - Applying SDT for Fitness

Products that are used, retained, and grown organically require building something that satisfies a user's needs or desires. Applied Behavioral Science in the digital space brings an additional toolkit to ensure that features encourage an end user to act in a new way. Ideally, these new actions will placate the desires or fulfill the user's needs in a way that increases user satisfaction and, in turn, positive business outcomes. As the vast nature of psychology would suggest, we are complex and filled with many different desires, goals, and pursuits. In some cases, these needs align with our desired outcome, such as when most individuals intend to be organ donors if they pass away unexpectedly. When this happens, a simple nudge, such as a default, can be utilized to align people's behavior with what they desire.

Newer research in Applied Behavioral Science suggests these nudges as a panacea are only applicable on occasion. Instead, these low-hanging fruits have already been discovered and corrected in many cases. Those who create interventions are left with the problematic realism that human behavior cannot align with individual passions at scale. This introduces why the personalization of experience to maximize user outcomes is necessary.

I have written about the importance of personalization and how a team can put themselves on the track to tailored interventions. However, I want to illustrate this process with a detailed example. As a reminder of what personalization of interventions looks like at a digital company, I will reference a favorite case study presented by Cedar, where they clustered groups based on their bill-paying patterns. With these clusters, they now had behavioral data-informed groups to build features targeted towards.?

Benefits of Personalization?

On top of building a product that works for users, ideally yielding higher levels of satisfaction and purchase, personalization also helps alleviate other common problems that product teams face. The first is not catching unintended consequences when only measuring usage. Customers are becoming more informed about the outcomes that using a product or service causes them in the short term and their everyday life. To achieve personalization, a team must measure the impact of their features and, in doing so, will be able to identify when certain groups interact with specific features that there is a risk of downside. These features can then be hidden or deprioritized.?

Second, by utilizing personalization, a company can avoid feature bloat even as it releases new features. How often have you opened an app and been overwhelmed with the novel routes you could explore? You are not alone; as teams continue to grow what they offer, it is uncommon for them to deprecate features that may not be widely used anymore. In a world of personalization, the product only shows users the features most likely to help them achieve what they desire, avoiding choice overload.

Making it Real: Peloton Example?

While all of this may sound exciting and be something you desire to implement, it is still challenging to determine where to begin. Thus, the remaining writing of this article will focus on a case study of how to apply a foundational theory to a product and develop a personalization framework. The hypothetical example we will utilize is Peloton, although one can replicate this approach across any Fitness application.?

Peloton bills itself as an effective way to achieve fitness goals through physical equipment and a robust digital experience. Users come to it, as with many digital fitness applications, wanting to attain some fitness goal. These desires significantly differ; some people are embarking on a lifelong journey to lose significant weight and improve their overall physical health, others want to drop a few pounds before a wedding, and others are already in great shape and looking to maintain it or push themselves even further. Thinking through each of these desires, you can imagine how important it is to expose different environments to each group, including the content presented, the intensity encouraged, the metrics shown for progress, and the type of feedback delivered when they do not hit their goals; to name a few elements.?

Creating an experience tailored to an individual not just at the surface but in how it directs them to act to match their goals is the point of personalization. To manage these differing desires and expectations, one must find an academic or professional framework to plot individuals into a category. Each use case will differ, but to identify the correct model, one should investigate meta-analyzes that academics in the associated fields have done along with the underlying themes of popular press books. For fitness, doing a cursory analysis of writing about how to approach fitness habits at scale, one will find that the self-determination theory is one of the most researched frameworks.?

Introduction of Self-Determination Theory Applied to?fitness

Self-determination theory (SDT) is a psychological framework focusing on the underlying motivations driving human behavior. It posits that individuals have three innate psychological needs: autonomy, competence, and relatedness. When meeting these needs, individuals are more likely to be motivated to engage in and persist in activities that are important to them. SDT is highly applicable across various domains of human behavior, including education, work, relationships, and health

In the fitness context, SDT suggests that individuals who feel in control of their exercise choices believe they can achieve their goals and have a sense of connection to others in their fitness community are more likely to engage in regular exercise and maintain their fitness goals over time. By providing opportunities for choice, fostering feelings of competence, and creating a supportive fitness community, fitness programs can help individuals meet their psychological needs and increase their motivation for long-term fitness success.

Within this is a framework called the self-determination continuum (SDC). This framework describes the different types of motivation individuals may experience when engaging in an activity specific to us, fitness. At one end of the continuum is amotivation, where individuals lack any motivation to engage in an activity and may feel disinterested or indifferent. Next is external regulation, where individuals engage in an activity purely for external rewards or to avoid punishment. Moving along the continuum, introjected regulation involves individuals who engage in an activity due to internal pressures or a sense of obligation. The identified focuses more on doing an activity because it aligns with an individual’s identity. Next, identified regulation involves individuals who engage in an activity because they value its benefits and align it with their personal goals. Finally, intrinsic motivation represents the most autonomous form of motivation, where individuals engage in an activity because they find it inherently enjoyable or satisfying. Users at each stage of this continuum will engage in exercise for different reasons and thus be motivated by differing messaging, metrics, and incentives.?

Moving the framework from theoretical to application

The first personalization step is completed with an established underlying framework to target different users based on their desires and motivations. Now, there are a few other resources that you need to do this at scale:

• Interventions: A collection of interventions that can be deployed to users. Examples of this would be different messages to be sent as notifications, different metric dashboards to showcase relevant progress, and full features which may be relevant to only some users, such as social leaderboards.?
• Experimentation Platform: A system of deploying these interventions and testing their impact on the users. This is generally done utilizing a feature flagging system (Ex. LaunchDarkly) combined with an experimentation platform (Ex. Optimizely)
• Data Collection: A back-end system is in place that collects the necessary behavioral data to enable the learning of user group preferences and the ability to deploy different experiences to users based on their group membership.

Starting from scratch, Peloton would utilize the data they currently capture on their users and deploy onboarding surveys or other attitudinal data collection methods to build a basic profile on each user. They would place users into one of these SDC groups with that data. For example, a user who reports choosing cardio to lose weight after their doctor suggested they should be placed into the external regulation category. They would then get messaging focused on the rewards attached to achieving rewards from completing workouts. Meanwhile, someone who reports they work out because it is a part of who they are would be placed in the integrated regulation cohort.?These users, in turn, would be sent messaging that reminds them that fitness people engage in exercise every day.

The researchers must then validate the two main assumptions with users placed in initial groupings. First, the team must confirm that the characteristics of an individual match the correct SDC group they have in place (group membership assumption). The second is that the proper interventions are deployed on the right SDC group (targeted intervention assumption). Testing these assumptions is the starting point of personalization and where it is necessary to have a solid experimentation platform in place.?How to test:

• Group Membership Assumption: This will be a game of constantly refining the model which places a user into a cohort. The team will use the initial training data to identify behavioral data patterns, which can predict future users' group memberships. As the groups grow, the team must collect further attitudinal data to validate that the model accurately places users. In these tests, the Peleton team would want to survey what motivates the users and then bucket them accordingly to test if their model shares the same results.
• Targeted Intervention Assumption: With initial group membership established, interventions should be deployed across the populations and measured according to SDC group membership. Here the researchers would be looking for which interventions have an outsized effect on certain groups. As differences in the effectiveness of the interventions begin to emerge, they can be deprecated for those where it doesn't work and highlighted for those whom it benefits most. An example of this would be a social leaderboard that might be useful for someone motivated by group membership and identity reminders but counterproductive for someone just trying to get a job (their workout) completed.?

While I have described the types of studies that must be conducted at a high level, significant engineering, and data science requirements must exist to accomplish this. That is out of the scope of this article, but I will point you to a few resources to read further on this. Resources: Uber Engineering Blog; Launch Darkly Blog; Optimizely Blog; and Shopify Engineering Blog (Keywords: Experimentation, Testing, Infrastructure)?

As described, this personalization approach is a surefire way to get users an experience that is most likely to cause change for them. It also gets a team on the path towards proper, person-level tailoring that can occur after the above assumptions are tested repeatedly, allowing for more groups and interventions to be created. This said, it is not easy and undoubtedly resource intensive. Thus, all of this must connect to a broader business goal. Before launching any of this work, it may be adventitious to run a simple survey or conduct a simple analysis to make the correlation between user use case fulfillment and positive business outcomes such as up sales and retention. If these cost savings and revenue-increasing metrics improve through this work, it can begin to pay for itself.

This detailed article is intended to make the future of the personalization of interventions that I write about more real. I have combined my previous work with a theoretical example that is intended to be replicable. While no one is doing this at scale in a significantly successful way, one of the leading groups is Lirio, who is doing this in the healthcare space. Another is Humu in the well-being and productivity space. Lastly, another company to watch out for from a pure marketing standpoint is Mutiny, which creates tailored company landing pages.?

If you find this vision useful or interesting, please let me know. If you know of any other companies working on this effort, I would also like to hear more as I try to keep tabs on the leaders in this space.