Barriers to building behavior change products
Why we struggle to help people exercise more often, save more money, and make other changes that matter
In the past few years, building products in healthcare and finance, I’ve realized something:
Despite thousands of books, articles, and studies on the topic… we’re still early on in learning about changing behavior. We’ve found some marketing hacks that work in the short term (especially to sell things!), and some edge cases where a behavioral science technique can drive small improvement.
But for the big, fundamental behaviors that matter (diet, exercise, savings…) for any given individual… it’s hard to find the answer for you.
We know anecdotally that change is possible. A sudden health scare for you or someone in your family, a big life milestone that you want to save for — these things can prompt an epiphany. But short of those largely uncontrollable life events, we don’t have a widely accepted point of view on how to create change.
I see a few things that are holding us back:
1. Creating a common framework for the behavioral tools that we have to deploy
There are dozens of one-off books and articles looking at specific behavioral science tools, like “choice architecture”… and still more talking about generalized frameworks (“nudges” or “framing”). The way we’ve sliced and diced behavioral tools without creating a reliable framework or hierarchy makes it difficult to use them in applied contexts, like product strategy. Have we really considered all our options for what technique to apply here? It’s hard to say without a mutually exclusive/collectively exhaustive list.
The behavior change techniques taxonomy is a great starting point — but there’s an opportunity for us to further consolidate this taxonomy and bring it into industry use rather than academic obscurity.
2. Connecting existing behavior change results to appropriate context
There is a temptation in existing behavior change literature to assume that a given result is broadly generalizable. If a time limit helped more people get flu shots, it must also help them… go for a run, change their diet, save more money?
In addition to a taxonomy for behavior change techniques, we need to start categorizing the context (target user group, type of behavior) in which behavioral interventions have been applied. Social norming, for example, has been proven to have wildly different results when applied in different contexts. I have to imagine that the same is true for many other techniques. We have to start building a knowledge base for what to apply where, and when.
3. Understanding what behavior change techniques will work for you
Beyond understanding the techniques we have, and how to apply them in general, we still need to understand you to match you to a behavior change strategy. What motivates you, or your propensity for a certain kind of behavior, is not cleanly observable from the basic demographic information that most companies have.
I see a few big, longer-term opportunities here:
Expand the use of validated surveys. Many companies have attempted some flavor of onboarding survey to try and learn more about you, but these are often new, bespoke instruments with no known correlation to the behaviors they aim to predict. Validated instruments do exist in some contexts (Patient Activation Measure in healthcare, for example) but not in all, and are not widely used in industry. We will need to both create and test these.
Invest in machine learning. In the context of ads, ML seems awfully effective at learning those soft, subtle things about us — what we’re interested in buying, what colors, what textures, what type of imagery appeals to us to get us to click and take just one more step. That subtle, persuasive power could be applied to help us make progress down the funnel of healthier habits rather than the funnel of e-commerce purchases.
Make better use of genetic data. In Blueprint, a leading behavioral geneticist reports that “genetics explains more of the psychological differences among people than all other factors combined.” In order to understand the “soft stuff” about you — whether you need a ‘tough guy’ weight loss coach or a ‘cheerleader’, whether you need to have complete control over your savings, or whether you might be receptive to quiet automatic transfers — we need data. Some of this can come from self-reports, or from tracking of past behavior (like I mention above) — but assuming a future where gene sequencing continues to cheapen and spread, and our interpretation of the results continues to improve, I see an opportunity to use it, with permission, in the careful matching of people to the behavior change path they might find most supportive.
Together, these tools could feed each other. Based on your readiness to change today (measured via a validated survey), we could select an intervention most likely to support you and your unique challenges (based on your genetic data paired with current behaviors), presented to you in a way that compels you to take action (learned by ML algorithm based on your previous actions). We could learn a relationship between your genes and your actual behavior, and figure out how much to rely on genotype vs. phenotype. We could build a system that takes in what we know (self-reported, genetic, behavioral data), weights the different elements appropriately, and powerfully targets personalized behavior modification strategies.
While there are many companies currently touting their expertise in behavior change, I find that what they’re usually touting is the human touch. Failing to act on any of the above, it’s still very possible to use human connection (motivational interviewing, etc) to suss out the right intervention for a person. Using human coaches (whether financial advisors or healthcare providers) has a scaling problem, and can only take us so far. I’m excited to work on these challenges that will fundamentally enable more scalable behavior change in the areas that really matter.
