Six Prescriptions for Building Healthy Behavioral Insights Units

Over the past few years, we have had the opportunity to work with over 20 behavioral units as part of our Behaviourally Informed Organizations partnership. While we as a field know a fair bit about what works for changing the behavior of stakeholders, what can we say about what works for creating thriving behavioral units within organizations?

Based on our research and hard-won experience working with a diverse set of behavioral units in government, business, and not-for-profit organizations, we have seen many success stories. But we have also seen our share of instances where the units wished they had done things differently, units with promising pilots that didn’t scale well, units that tried to do everything for everyone, units that jumped to solutions too quickly, units too fixated on one methodology, and units too quick to dispense with advice without thinking through the context in which it will be used.

We’ve outlined six prescriptions that we think are critical to developing a successful behavioral unit—three don’ts and three dos. We hope the advice helps new and existing behavioral units find their path to success.

Prescription 1: Don’t anchor on solutions too soon

Many potential partners approach behavioral units with a preconceived notion of the outcome they want to find. For instance, we have been approached by partners asking us to validate their belief that an app, a website redesign, a new communication program, or a text messaging strategy will be the answer to their behavior change challenge. It is tempting to approach a problem with a concrete solution in mind because it can create the illusion of efficiency.

However, it has been our experience that anchoring on a solution constrains thinking and diverts attention to an aspect of the problem that might not be central to the issue.

For example, in a project one of us (Dilip) was involved in, the team had determined, very early on, that the most efficient and scalable way of delivering their interventions would be through a smartphone app. After extensive investments in developing, piloting, and testing an app, they realized that it didn’t work as expected. In hindsight, they realized that for the intervention to be successful, the recipient needed to pay a certain level of attention, something for which the app did not allow. The team made the mistake of anchoring too soon on a solution.

Early anchoring of the app as the answer to scaling meant that the team did not question (or test for) the appropriateness of an app as the carrier of the interventions. They also fast-tracked the process of trying to understand the “first principles” of the problem.

To counter this solution-fixation, we advocate for the use of a My behaviour change challenge (myBCC) statement, a comprehensive user-journey mapping, a thorough investigation of possible frictions to the desired behavior, and finally, a How might we (HMW) statement. Templates for the myBCC and HMW statements are below. It is only after articulating the behavior change challenge that we then recommend ideas for how the unit should start thinking about potential solutions.

Prescription 2: Don’t simply take things off the shelf at the nudgestore!

It is tempting to read about successful use cases and design interventions that mimic them. We refer to this as “shopping in the nudgestore.” Understanding what others have done is a good start to solving your own problem, but don’t assume what worked in their specific context will work in yours. Doing so can be counterproductive.

Given that decision-making and behavior are influenced by supposedly irrelevant factors, it is important for units to scrutinize the use case or published paper and ask 1) is the situation in the use case similar to my situation? And 2) are the participants in the use case similar to my participants?

Elements of the situation might include:

  • The physical environment (where decisions were made and actions taken)
  • The social environment (the presence or absence of social factors like people or institutions with whom they interact, or other cues like occasions and surroundings)
  • Time (the time of the day, day of the week, the time-pressured experience by the end user)
  • Attention (whether the intervention forces the end user to pay attention to it)
  • The medium through which the intervention is delivered

Changes in any of these elements might change the outcome and the success of the intervention. For example, the effectiveness of reminders depends on factors such as when they are sent, how they are delivered, and how long an attention span is needed to act on them.

If you determine that the situation or participants are significantly different, we recommend nothing less than gathering your own evidence through surveys and interviews, experiments, and ethnographic techniques to increase the confidence that the translation might be successful.

Prescription 3: Don’t rush into behavioral solutions

While most challenges in organizations relate to behavior change, the solutions to these challenges might not always be behavioral in nature.

For instance, systemic issues (like overly complicated processes or confusing information) might be the reason end users do not display the desired behavior, not because of any psychological considerations. If processes are inherently sludgy, no amount of psychological interventions will help.

Before thinking through any behavioral interventions, we recommend that units work with all responsible parties to ensure that processes are as efficient as possible, and that the behavioral pipelines are kept clean and sludge free.

Prescription 4: Do develop a clear positioning statement for your unit

A strong positioning statement is a critical piece of the go-to-market strategy, yet many behavioral units do not have one. We’ve observed that there is a naive belief that the science will be effortlessly embraced by other stakeholders. But, as we’ve learned through our research and experience, the value that we create might not be apparent to others.

The best units have a clear statement of what their value to the organization is. They clearly articulate how they are different from other units that might be seen as being behaviorally informed (like marketing and sales or market research), what can they do those other units cannot, and who their target audience is within the firm. A clear positioning statement helps units remain focused on where they can add the most value, communicate their value to stakeholders, assess the impact of their work, and pivot if needed. They are also helpful in determining noncore activities that the unit should not be spending too much time on and looking for ways to deliver value more efficiently.

Units can write positioning statements for different products or activities that they offer (see sample positioning statements below). The appropriate positioning for a unit will be different in different organizations as a function of what other capabilities they currently have and what their primary value add to the organization is.

Prescription 5: Do designate a heterogeneity and a scaling advocate in the team

We know from recent research that the effects of choice architecture and interventions are often heterogeneous—they differ for different groups of end users—and that they might not successfully scale.

For instance, an intervention to make pension statements in Mexico simple and engaging mildly increased contribution rates on average. However, the average mild result came from two sources—people whose pension funds were performing particularly well showed a strong positive effect, while those whose funds were performing relatively poorly showed a negative effect. Clearly, a different intervention might have been better for the second group. Likewise, a text message about family security worked mildly well on average, but again showed strong positive effects for people aged 28–42 (presumably people with families) but backfired for younger consumers (presumable single or without children). Again, a different intervention might have been better for this second group.

However, part of the challenge is that heterogeneity and scaling are usually not thought through during the early phases of a project, when a design is being developed and tested. Additionally, the mandate of most behavioral units does not include scaling, which is often considered to be the responsibility of another department.

We recommend having at least one team member responsible for thinking through and identifying possible heterogeneous effects and challenges to scaling. And we recommend identifying this person (or people) as early in the process as possible. Doing so will help the team better anticipate, design for, and respond to challenges that could undermine the aims of the project. Thinking about heterogeneity and scalability early can save help save time, money, and energy, all scarce resources that businesses and governments could do without wasting.

Prescription 6: Do use multiple methods

We recommend that units use multiple methods to collect evidence on the efficacy of their interventions. This includes explanatory methods (e.g., observation, interviews), A/B testing, laboratory experiments, randomized controlled trials (RCTs), and data science in conjunction with RCTs.

RCTs are thought of as the gold standard of evidence. While we don’t disagree, we have also learned that a fixation on RCTs can lead to the neglect of other valuable methods, or units simply giving up on an intervention once they realize that an RCT isn’t feasible.

Explanatory methods such as observation, journey mapping, interviews, and the analysis of archival data are useful for identifying problems and generating solutions. The broad lens and user-driven view taken by these methods allows the unit to identify a broad range of frictions to existing behavior and helps teams avoid solution-mindedness.

Methods such as laboratory or online experiments might be useful to shortlist potential solutions or iterate towards better ones. These techniques might involve hypothetical choices but can be done quickly and cheaply. Methods such as RCTs or the econometric analysis of observational data sets might help to determine the efficacy of the intervention, but given their complexity, cost and time, they should be used as the final piece of evidence before scaling the solution.

Crafting a successful behavioral unit, like a crafting an intervention, depends on understanding the context, what’s worked before, and what the unit needs to accomplish then charting the appropriate course. We hope these six prescriptions help new and current units do just that.

Disclosure: Dilip Soman and Bing Feng are member of the Behavioral Economics in Action at Rotman (BEAR) Research Center, which provided financial support to Behavioral Scientist as a 2023 organizational partner. Organizational partners do not play a role in the editorial decisions of the magazine.