Note: While reading a book whenever I come across something interesting, I highlight it on my Kindle. Later I turn those highlights into a blogpost. It is not a complete summary of the book. These are my notes which I intend to go back to later. Let’s start!

  • What made the iPhone work is really two whys that together bridged what psychologists call a counterfactual world—the world that doesn’t exist but could—with the one we live in. Apple imagined a world in which people used their phones everywhere, all the time, to honor a vast range of motivations. Then it built a device that guided people to do that, not by “challenging the status quo” in every aspect but by asking “Why would people want to do that in the first place?” and “Why aren’t they doing it already?”

  • Behavior change; when you do it successfully, you remove the inhibiting pressure of doing it first, so others frequently attempt to fast-follow, though without any understanding of the validated pressures you used to create the behavior change and thus a significant handicap on their success.

  • The Intervention Design Process exists simply to make such alternative realities come into being, to take potential insights and run that potential to ground to see if it can create value. There are four major types of potential insights: quantitative, qualitative, apocryphal, and external.

  • The first is driven by, as the name implies, data. It usually comes out of either the recognition of a pattern—like an unexpected and unexplained correlation that seems to keep popping up—or the study of outliers, either positive or negative. This is one of the reasons that simply wallowing in data is so important and why not all data wallowing should be hypothesis driven. Finding novel potential insights is about noticing something that hasn’t been noticed before, and that’s awfully hard to do if you’re relying on existing hypotheses to guide you. You become your own worst limitation. When you let the data guide you to a potential insight, you often discover things that you feel like you’ve known all along (because your brain likes to feel congruent and smart) but that you would never have generated a priori.

  • Qualitative insights are similar but derived from subjective experience rather than carefully curated numerical tables. If you’ve ever been people watching and had that little tickle in your brain that says, “Hmm, that’s interesting,” you’ve had a qualitative insight. And the best way to produce those is by talking to and observing the diversity of people in the world (whether or not they already use your product; sometimes it even helps to intentionally learn about people who don’t), something we all know but rarely do. Charles Pearson, one of the user researchers at Clover Health, organized volunteer trips to senior living communities for Clover employees so that they could see what it meant to be an older adult in America, and that’s exactly the right strategy; you can’t force insights, but you can create an environment in which they are more likely to happen. The quantitative equivalent is having an open data approach to your organization’s data warehouse, another thing Clover does well: anyone at the company can learn thirty minutes of SQL and then start generating insights.

  • Apocryphal insights aren’t directly observed; they just seem to be common knowledge in your organization. Bing in the Classroom really started here. Everyone at Microsoft just knew students weren’t searching, without really being sure why they knew that. Pay attention to apocryphal insights, especially when you first come into an organization. I have an informal rule that I won’t manage people the first year of a new job, and it is precisely because I want to hear what everyone thinks they already know so that I can start to validate or discard it. And when I do manage people, I pay for newbies to have lunch with colleagues from across the company and document the apocryphal insights they hear for future validation. Because once you’re immersed, you just tend to accept what is believed by others as actually known and lose that outside perspective all too quickly. Finally there are external insights. These come from beyond your organization, out in the wide world. Research papers are good for external insights, but so is simply cross-pollinating with other industries and disciplines. One of my favorite sports is taking random grad students to lunch and then just trying to figure out what everyone in their field believes to be true that might be applicable to my work. Grad students are the single most untapped resource I know of in academia. They’ve chosen to devote their lives to a topic, but nobody in industry ever gives them the chance to talk about it. Take a note: buy lunch for grad students. Better yet, use them as consultants who actually know something and pay them fairly for their knowledge. Because as with apocryphal insights, you need input from people other than you to really get the most out of the IDP.

  • No matter which kind of insight gives rise to the process, you should never simply assume truth; remember, in science everything is assumed false until proven. Particularly with apocryphal and external insights, people will speak with confidence based simply on something they heard from someone else, in some insane game of telephone that usually butchers whatever the original insight was. So once a potential insight is surfaced, it must be validated through insight validation. What we’re looking for here is convergent validity: evidence from diverse sources that supports the same conclusion. For example, if we are looking at the data around prescription drugs and suspect, from looking at home and fill addresses, that people might not be going to the optimal pharmacy, we should use other sources of insight to triangulate that as a potentially valid insight. For qualitative validation, we might talk directly to members, look at call transcripts and do call listening, or ask pharmacists what they’ve observed. Apocryphally, we can ask knowledgeable members of our organization and see if the insight rings true for them, totally subjectively. And we can look at external research (Google Scholar is your friend) on pharmacy selection trends to see what is known beyond our organization.

  • Validation must be deeply embedded and constant throughout the entire IDP. Think of it like building a table—you want multiple legs, as far apart as you can get them, to hold up your conclusions. This is how we resist the Mad Men world, in which people build things based simply on their own personal beliefs and then manipulate data or other sources so they appear to support it (data scientists everywhere are nodding knowingly right now). Your brain, being the lazy mofo that it is, has a tendency to cheat and use something we call the confirmation bias: as soon as you start to believe something, it begins to selectively attend to evidence in order to support that belief, because changing your mind costs cognitive resources and your brain is a couch potato. The more varied your sources of convergent validity, the harder it is to fall victim to confirmation bias. One effective way to get the legs of your table farther apart is to assign each type of validation to a different kind of researcher who specializes in each method, then cross-training them to be able to check one another. If each researcher works independently before they come back together, there is less of a tendency to cheat and proceed with an unvalidated insight or to reach group consensus too early. My team at Clover Health has both quantitative and qualitative researchers, as well as a rotating three-month fellowship for an outside master’s or PhD student to do nothing but external validation (once again, grad students are your friend). Once a week the researchers and project managers come together to compare insights for convergent validity and to find potential insights generated by one discipline that the others can dig into.

  • Good managers reward invalidating an insight as much as validating one.

  • It is also important to focus on what you are validating. Too often people use research simply as a post hoc process of confirmation bias to affirm whatever they already think they know, particularly when there are differences in organizational power between those who suggest the insights and those who are tasked with validating them. This is exaggerated in a “just ship it” culture, where the minimum viable product (MVP) takes the place of user research. The idea becomes that if you simply launch a product, people’s reaction to that product is the only research you really need. But what are you truly validating? If we simply launched Flamin’ Hot Cheetos and they weren’t immediately popular, what should we conclude? That the Latinx market is uninteresting? That the snacks were marketed badly? That the flavor isn’t right? And if they are popular, how do you launch the next version when you have no idea why the first one worked? The point of validation is that it allows us to build toward behavior change not by popping off a silver bullet and hoping we hit the target but by a scientific process of advancement that makes sure we do.

  • Generating insights horizontally—that is, agnostic of hierarchy—throughout a company works for many reasons. With Richard, there was no formal suggestion process at all: he had an idea and, despite his job title, ran with it all the way to the top. It’s a rare but exemplary instance of seeing the overlap between your consumers and your employees, and in this case it was a win-win all around; Richard is now an executive at Frito-Lay and a motivational speaker, spreading the word about the value of each person having a voice in their work. The tendency to focus on processes over outcomes is actually a natural psychological operation: because we spend the majority of our cognitive resources on what is happening right now—which makes sense, since we do actually have to act in order to create change—we’re biased toward focusing on those immediate actions and how we’re doing them, rather than on their outcomes. Means over ends, or whats and hows over whys. The actions are more psychologically available, while the outcomes recede into the mental distance. That’s why frameworks like the IDP are so important: they help us make better, more clear-eyed choices about how to get things we want by imposing a process that fights our natural biases. At its core, a behavioral statement is simply a set of binary conditions that can either be satisfied or not. A typical behavioral statement has five variables that come together into a single sentence: When [population] wants to [motivation], and they [limitations], they will [behavior] (as measured by [data]).

  • Population = the group of people whose behavior you are trying to change Motivation = the core motive for why people engage in a behavior Limitations = the binary preconditions necessary for the behavior to happen that are outside your control Behavior = the measurable activity you want people to always do when they have the motivation and limitations above Data = how you quantify that they are doing the behavior Note that each of these things can be answered with a 0 or a 1, a yes or a no. You are either in the population or not, you either want something or don’t, you either meet the conditions or don’t, you either do the thing or don’t, and there’s either evidence or there isn’t. How do behavioral statements work in the wild? Let’s go with one of the clearest examples I know: Uber has done a remarkably good job of driving a very clear, direct behavioral statement, and that is a large part of why it is successful. Quick aside: fuck Travis Kalanick. If you are an exec and you aren’t actively fighting sexism in your own company, you’re a sexist. And even though he isn’t the CEO anymore, he still owns a large portion of the company and every Uber ride enriches him. I use the Uber example because I want a particular behavioral outcome (you apply the IDP after reading this book), and you don’t always have to like an intervention for it to work. But seriously . . . fuck Travis, take a Lyft. Uber was founded to solve the problem of getting around in San Francisco. Internet companies were flourishing, but unlike the business hub of New York City, San Francisco lacked a massive subway system. Uber was founded to solve the problem of getting around in San Francisco. Internet companies were flourishing, but unlike the business hub of New York City, San Francisco lacked a massive subway system and the ability to simply step out into the street and raise your hand to get a taxi. The growth of an industry drove a new, strong motivation, and thus Uber’s initial behavioral statement might have looked something like this: When people want to get from Point A to Point B, and they have a smartphone with connectivity and an electronic form of payment and live in San Francisco, they will take an Uber (as measured by rides). Seems simple, right? That’s the advantage of a good behavioral statement: when written with care, it is easy to digest, powerfully clarifying, and grammatically correct (here’s looking at you, dear copy editor). The individual conditions carry little risk of misinterpretation, are binary, and can be measured. But the fact that a behavioral statement is simple in expression doesn’t mean it is easy to write one. Let’s break down Uber’s statement to see why it works. POPULATION = “PEOPLE” Uber really was an app for everyone. If a chicken figured out how to use a smartphone, I’m pretty sure Uber would have given that bird a ride across the road. This is actually fairly unusual on two fronts. First, most products and services have an audience for whom they are right and a much larger audience for whom they’re not, because it is hard to find a universal motivation that allows for a universal population. Second, the general rule of thumb is that the fewer resources you have as an organization, the narrower and more specific your behavioral statement needs to be. Uber was the epitome of a go-big-or-go-home bet; it wasn’t prepared to settle for just being a better version of your local cab company. So an initially very broad population worked for it (but has and will for very, very few others). For Richard Montañez, the population in his behavioral statement was Latinxs. For me, working on Bing, it was K–12 students. MOTIVATION = “GET FROM POINT A TO POINT B” Uber also had the benefit of an easily defined motivation with a few special characteristics. First, the desire to go from A to B was fairly general. The company’s services didn’t have a specific population, as noted above, but they were also broadly applicable across time and location. Certainly there were surges to deal with during rush hour and dips in activity at 3:00 a.m., but people need to go somewhere every single day of the year, rain or shine, from many Point A’s to many Point B’s. Uber’s need was also general in another sense: people were already habituated to using a diverse set of methods to get around. If you lived in San Francisco at the time, you were likely using a combination of cars (including taxis), trams, trains, buses, ferries, and walking. It wasn’t hard to get people to adopt a new method of transit, because they were already using so many. Certainly there were habits ingrained for utility’s sake—people knew the most efficient or scenic or safest ways to get around the city, depending on what mattered to them—but these habits were weak and easily disrupted, because the biggest habit was using many methods of transport. Taken together, these were deceptively powerful advantages for Uber in its early stage. Picking the right motivation isn’t often discussed but, done well, it can be incredibly helpful. LIMITATIONS = “HAVE A SMARTPHONE WITH CONNECTIVITY AND AN ELECTRONIC FORM OF PAYMENT AND LIVE IN SAN FRANCISCO” This was actually the most daunting part of Uber’s behavioral statement. In 2009, cell phone ownership wasn’t ubiquitous, cellular service wasn’t a given, and people weren’t as used to storing credit card information in an app. Hell, 15 percent of the U.S. adult population didn’t even have a form of electronic payment in 2009.4 But as a startup, Uber didn’t need everyone to make it work. It just needed to demonstrate a viable enough model to attract the next round of funding, and a young, tech-centric city like San Francisco was exactly right. Although all of a behavioral statement’s variables are binary, it’s especially worth noting that limitations don’t exist on a scale: they’re either yes or no, 1 or 0. Someone does or doesn’t have a smartphone, does or doesn’t have an electronic form of payment, and does or doesn’t live in San Francisco and these factors are explicitly outside the company’s control. This is important because it is easy to accidentally list inhibiting pressures as limitations. Being able to afford an Uber, for example, is really about expense, and perceptions of expense are variable: they can be strengthened or weakened by interventions. It is also important to be careful not to include limitations that you intend to modify through behavior change. Because while having the Uber app is a precondition for using Uber, it isn’t a limitation because it is Uber’s job to make that happen, whereas Uber wasn’t relocating users to San Francisco or giving them cell phones or credit cards; limitations exist precisely because we are explicitly choosing not to target interventions at them. BEHAVIOR/DATA = “TAKE AN UBER”/”RIDES” This is where Uber really nailed it: they knew exactly what they needed to get people to do (take an Uber) and exactly how to measure it (rides), with the added bonus that their data was generated by the product itself and didn’t have to be separately gathered. Certainly there were all sorts of other metrics, like sign-ups and app opens and all the other things that get tracked inside a company, but the behavioral goal itself was automatically measured, data-complete, and nearly immediate.

  • Writing a good behavioral statement is hard, one of the hardest parts of the IDP, and there are a couple of common pitfalls you have to avoid to do it well.

  • Easily the most common mistake is not being thoughtful enough about the actual behavior you want to change. This often comes because you’re focusing on sounding good instead of being good, trying to write a vision statement instead of a behavioral one. For example, for many years the Microsoft vision statement was “A computer on every desk and in every home running Microsoft software.” And while I love that vision and the great good it created (and still creates) in the world, it is a singularly bad behavioral statement. Why? Because the mere existence of a computer isn’t actually a behavior, or if it is, it is only the limited action of a one-time purchase. To really see the problem, imagine a world where the vision statement is literally true: every home and office has a computer running Microsoft software. Now envision them all unplugged, covered in a healthy layer of dust and sprinkled with last week’s laundry, because nobody actually wants to use them. Nothing about Microsoft’s statement is incompatible with that hypothetical reality, because the statement is about the existence of an object in a certain location and nothing more. It fails to address the actions implied in the vision statement coming true—those measurable behaviors that we know and love. You might laugh at this example, but it is actually a real one that led Microsoft significantly astray. For years, the Office product and tech teams were oriented toward sales as the metric that mattered. Accordingly, they focused on features that no home consumers cared about but that appealed to niche corporate clients. And so they introduced endless features focused on expanding the reach of the software into increasingly niche populations, making Excel the powerhouse behind modern financial analysis by creating a deep macro system and Word an essential part of the publishing industry by creating sophisticated markup languages. Don’t know what those features are? There is a reason for that—they probably weren’t built for you, but rather to sell an extra few licenses to some corporate customer. Hell, the salespeople were even paid commissions on the number of licenses they sold to enterprise customers. Because remember, if we want every single computer running Microsoft software, workplaces are the primary buyers of computers and so by orienting everyone toward sales, Microsoft could bring its vision to life. The problem was that when contract-renewal time came, CTOs were constantly trying to trim down the contracts because they had discovered that nobody in their organizations was actually using the software! The reason Google Docs came into being was that Microsoft wasn’t paying attention to the actual experience of using its software—the behavior it should have been monitoring. And when it did, pivoting internally from a sales metric to a usage metric (even for the sales team!), it created Office 365 and the ribbon bar, which made Excel and Word reasonable for those of us who aren’t in finance or publishing.  Potentially worse than picking the wrong behavior is picking no behavior at all, and that’s the second-most-common mistake. It also tends to happen when people focus on vision statements over behavioral ones and typically comes from marketing- or product-focused CEOs, often in the form of something insanely trite like “Our job is to make the customers love our product.” What the fuck does that mean? Love isn’t a behavior. You can’t physically observe it and thus you can’t measure it, so if you try to design for it, you will inevitably end up in the same boat as 2000s-era Microsoft. Customer love is just another post hoc rationalization for Mad Men to spend money on something they love. And because there is no way to measure customer love, there is no way to prove that any particular intervention actually makes a customer love your product. Which means that there is no way to compare interventions against one another, short-circuiting the whole IDP. Your marketing team will spend millions and justify it with customer love and at the same time your product, tech, sales, and other teams will be running in opposite directions with precisely the same justification. A statement without a behavior is a North Star you can’t navigate by.

  • Uber’s behavioral statement described a world in which people would always use Uber when they needed to go from Point A to Point B, not sometimes but always. As is true of other things in Uber’s history, that’s a ballsy statement. There are so many modes of transportation these days, and cars in general aren’t always the best option; there’s traffic, greenhouse gas emissions, potential parking fees, stolen tires, etc. All those limiting pressures are the reasons people don’t already choose a car all the time. So why describe the absolute? Because it increases the likelihood you’ll get there. A typical process-based design system asks: “If we are here, how are we going to move the ball down the field, closer to the goal?” I want you to ask: “In the perfect world the ball is already downfield in the goal and we have won the game. What play do we need to run to cause that to be true?” These two statements sort of sound the same because they both see the current world and desired world on opposite ends of the playing field and both try to get us from here to there. But this is where one of those pesky mental heuristics gets in the way: anchoring and adjustment.

  • The last mistake is a bit more subtle: refusing to evolve a behavioral statement. Because sometimes market forces shift and you have to pivot the behavior that your company is trying to create. And even with the most audacious goal, you sometimes actually accomplish it and need to broaden. Take Uber. You’ll notice I used the past tense as I walked through its behavioral statement. That’s because that statement has changed significantly since its inception and will continue to as the company evolves. That’s not unusual, especially for companies that are relatively new; iteration and scale both generally allow you to broaden your behavioral statement considerably. At the time I’m writing this, the Uber behavioral statement might sound something like this: When people want to get something from Point A to Point B, and they have a device with connectivity and live in a metro area in most countries, they will use an Uber (as measured by rides). There are some big changes in there to unpack. The first, and arguably most important, is that Uber has expanded from a local transit company to a logistics company. It became so successful at acquiring drivers that it couldn’t create enough demand, spread out over enough time, to fill their available drivable hours. So it started delivering more than people: takeout, groceries, and anything else that needs door-to-door transportation, thus expanding the motivation from moving just people to moving anything. That’s huge.

  • It opens up entirely new markets, allows the company to guarantee greater revenue to drivers by smoothing out the demand curve, and creates the potential for very different forms of partnership with other large enterprises. And it also makes Uber more resilient; if people go fewer places physically because of changes in pressures beyond Uber’s control, that will hurt the company’s bottom line. But if they’re not going places, people will likely need more things brought to them. Because Uber can now meet that need, it is better prepared to weather macrobehavioral shifts. And it’s indeed possible that a behavioral statement changes because of external shifts, rather than an internal shift like the growth of your product. People used to buy watches so they could know what time it was. Then cell phones were born and suddenly everyone knew the time all the time. Yet watches haven’t disappeared off the face of the earth or the wrists of people. Because when the watch companies saw everyone pulling out their phones to meet the need that their product once met, they evolved their behavioral statement and found a new need watches could fill: status. Today you see people wearing watches less to tell the time than as a way to display something about themselves. Timex-ers send a message that they value clarity and economy and nostalgia; Rolex-ers tell the world all about their black cards and black cars.  

  • At the highest level, behavior change is about interventions that move us from Point A (the world as it is) to Point B (the world as we want it to be). And if our insights describe Point A and our behavioral statement describes point B, what remains is to understand why Point A isn’t already Point B. That is, we need to map the pressures that create the distance between what we have and what we want, so we know what it is we need to change.

  • Promoting pressures—the up arrow—make a behavior more likely and inhibiting pressures—the down arrow—make a behavior less likely. What we actually do is determined by the net product of those forces. If the promoting pressures overcome the inhibiting pressures, we act. If the inhibiting pressures are stronger, we don’t. And both sides are equally responsible for the ultimate behavior: we can never say people don’t act because of a lack of promoting pressure, because it could equally well be phrased that the overwhelming inhibiting pressures are responsible.   

  • Picture one of those big Mylar balloons you got for your birthday as a kid. It’s full of helium and suspended in front of you, hovering in what we’ll call an inactive state. The promoting and inhibiting pressures are balanced, so nothing is really happening at Point A, the current state of the world. Now the balloon, as every storybook will tell you, really just wants to float up and up and up into the sky and fulfill its balloon destiny—that’s its happy ending, its Point B, its desired behavioral end state. If you give it a little bump from below (an additional promoting pressure), you disturb the equilibrium and overwhelm existing inhibiting pressures like gravity, causing a behavior change. If you wanted to make liftoff even more likely, you could add a wind machine underneath or a little more helium inside. But what if it were raining so hard that the balloon couldn’t rise? Or what if, while you tried to push it upward, I pushed down against you? To get the balloon moving up against those downward forces—the inhibiting pressures—you’d have to push much harder, maybe add a little propeller. Or you’d have to find a way to reduce the inhibiting pressures, by blocking the rain or pushing me out of the way or, if you were really ambitious, reducing the force of gravity itself. That is actually what we are trying to do by mapping the pressures. By understanding the rain and my pushing down and your pushing up and gravity and all the rest, we are laying the groundwork for creating the interventions that effectively change that behavior to get us to the world we want. We can’t start adding or removing pressures until we understand, at least broadly, what exists now.

  • M&M’s: Start with our up arrow. Why do we eat M&M’s? The easy answer is that they taste good. Taste is a powerful promoting pressure, which is why Mars has spent millions of dollars coming up with different M&M flavors, more than forty so far (including the very unchocolate chili nut). This is clearly foolish, as we all know that peanut butter M&M’s are the pinnacle to which all other M&M’s aspire. Yet Mars keeps gleefully pumping out flavors, simply because taste is clearly the main reason to eat M&M’s or any other candy. Yet it isn’t taste alone that has sold billions of M&M’s. M&M’s are beautiful. We have a basic attraction to foods that look good, and we are preprogrammed to love that array of strong primary colors (you can blame your addiction on an evolutionary preference for brightly colored fruits and vegetables). People will eat more M&M’s from a bowl that contains more colors,5 and in 1995, when Mars dropped the least vibrant of its palette (tan—you know, the poop-colored one), more than ten million people called in to vote for its replacement (blue!). Color may seem like a silly reason to eat M&M’s, and few people would call it out as important. Yet part of getting good at pressure mapping is recognizing that there is much more at play than what people consciously identify as affecting their behavior. That’s why we need insights and validations and why we’ll eventually run pilots; humans have very poor introspection into our own motivations. After all, you are theoretically a very logical grown-up who reads nonfiction books like this one for pleasure. Yet you still have a favorite color of M&M, despite knowing that there is no actual difference in taste among different colors.

  • One behavioral science trick that always helps with pressure mapping is flipping the scenario on its head and taking it to the extreme—imagining if the opposite conditions were true and thinking about how that would influence the behavior. Because a successful intervention is about creating a world that doesn’t currently exist, you’ll always need to be working through thought experiments and inhabiting imaginary realms. And sometimes those are worlds we explicitly don’t want, so that we can more clearly see the world we do. Imagine if M&M’s came in a shade of puke green. Or maybe piss yellow. Still think people would be so quick to shake the entire contents of a bag into their hand and toss them back like a human Pez dispenser (a candy I cannot come up with a promoting pressure for besides the dumb dispenser)?  

  • The trick to getting good at pressure mapping is learning to let go of your natural assumptions, to see the irrational and the counterrational as opportunities. It is also recognizing that there are diminishing marginal returns: no behavior can ever be fully pressure mapped, so knowing when enough is enough depends on the maturity of the market. Some categories have been around so long that focusing on the disruptive, unrecognized pressures is key; others are so new that just getting the most obvious pressures right is enough to produce real change.

  • The list of promoting pressures really could go on endlessly. M&M’s have strong positive cultural associations: they’re iconic, nostalgic, and distinctively American (indeed, the official candy of the White House). The brand is synonymous with lighthearted fun, as narrated by giant animated candies, an image Mars has spent hundreds of millions to create and maintain. They’re also ubiquitous—a regular option in the school or office vending machine that feels familiar and comfortable. Again, we wouldn’t eat M&M’s because of branding or ubiquity alone if they tasted or looked terrible, but given that they’re delicious, all of the extra cultural connotations certainly help drive consumption.

  • So we now know that M&M’s are delicious, beautiful, full of hunger-fighting calories, and coated with good feelings. You nodded along faithfully to all my previous paragraphs, like good readers, and are now entirely convinced about the many reasons we want to eat them. And yet you aren’t eating them right now. Boom! Talk about a counterfactual universe. You just agreed with that laundry list of promoting pressures and yet you live in a world where you’re not actually eating M&M’s most of the time.

  • So why aren’t you eating M&Ms right now? This is where my degree in mind reading psychology comes in handy. The reason you’re not is that you’re not sitting next to a giant bowl of M&M’s. How do I know? Because if you were, I can say with a high degree of certainty that you would be eating them. Physical availability is one of the key inhibiting pressures to M&M consumption and pretty much everything else in life; humans are remarkably attuned to proximity.

  • Physical availability is one factor, but so is psychological availability. In one office-based study, workers consumed an extra two chocolates per day when the candy was on their desk versus across the room, but they consumed two more when the bowl was clear versus opaque.

  • “Out of sight, out of mind” holds true after all. Google is famous for doing this at scale; it put all the office candy in opaque containers but kept the fruits and nuts in clear ones. Over seven weeks, employees ate a cumulative 3.1 million fewer calories. Calories are another example: promoting pressure in the form of blood sugar in the midafternoon, inhibiting pressure in the form of health concerns when I get out of the shower and am confronted with the dad bod. That’s why we have to be specific in a behavioral statement about not only the behavior we want but also the population and context in which we want it.

  • Take the strength of cost as an inhibiting pressure: a dollar may not seem like much, but watch a five-year-old save their allowance to buy a bag of M&M’s, or grapple with the fact that most of the world lives on less than $2.50 a day, and you may change your opinion about just how inhibiting a dollar really is.

  • What is more important than the specific pressures is considering both sides of the equation. Because it turns out, to borrow a turn of phrase from Dan Ariely, humans are predictably irrational. In a series of experiments in my lab, we showed that when asked to focus on creating more of a behavior, people almost exclusively generated interventions aimed at increasing promoting pressures, like rewards. When focused on creating less of a behavior, they disproportionately generated interventions ramping up inhibiting pressures, like punishments.

  • What Uber recognized is that the desire to go from Point A to Point B is sufficiently strong that the majority of the task is just reducing inhibiting pressures and Uber’s product and marketing are in lockstep on this framing. Sure, every once in a while it does a promotion like “Today the ride comes with puppies” or ice cream bars or flu shots, or it will be a Tesla. But what are the three emails you regularly get from Uber? “It is now cheaper than it was before” (lowered inhibiting cost), “There are now more drivers on the road” (lowered inhibiting wait time), and “We can now go somewhere we couldn’t before” (lowered inhibiting range). Uber’s entire business is based on reducing inhibiting pressure.

  • Indeed, one could argue that Uber’s strongest behavior-changing feature is automatic payment. Paying for a cab is a strong inhibiting pressure. The physical loss of money that marks a cash transaction is highly salient, such that even small children would rather pay with a credit card than cash to avoid that feeling of loss.8 In some ways, I’m disappointed that Uber was invented in San Francisco; if you’ve ever been in the East Village in New York City on a one-way street on a Friday, frantically trying to get the card reader to accept your credit card while a thousand cabs behind you lay on their horns, you have a sense of just how painful the process of payment, not even the cost, actually is.

  • Both promoting and inhibiting pressures can be used to change any behavior, which is why we draw the arrows and populate both sides to overcome our biases.

  • Once you’ve drawn the arrows and successfully avoided bias, where do all these pressures come from? In the Mad Men world you just make them up to justify the interventions you want to run. But in the IDP we generate them in the same way we generated insights: research and convergent validity. Fortunately, all of the interviews and data science you did in search of insights also naturally lend themselves to the mapping and validating of pressures.

  • Meetup was growing so rapidly that spam was increasingly a major problem. Marketers would create meet ups that were actually just sales pitches for their products, diluting Meetup’s interest-organizing behavioral goal. The team was designing interventions to combat spam when the CEO, Scott Heiferman, suggested adding a required checkbox to the meet up creation flow that read “I pledge to create real, face-to-face community.”

  • The golden rule of designing sign-up flows is that you don’t add anything unnecessary, because every additional step introduces greater inhibiting pressure and thus lowers the behavior that is registration. But he recognized that the checkbox also reiterated the company’s mission and that meet-up organizers are a passionate group; maybe they wouldn’t be deterred and the extra inhibiting pressure would be enough to reduce spam. And he did what is critical in intervention selection: he remained open to being proven wrong. Not only did the checkbox reduce spammers, but it actually increased the number of successfully created meet-ups by 16 percent. Because not only are meet-up organizers sufficiently passionate (promoting pressure) to overcome a checkbox (inhibiting pressure), but reminding them of that passion created additional meaning and actually strengthened the promoting pressure.

  • You can’t test something and then pilot it. With each advancement, you gain certainty about the ability of the intervention to actually create behavior change and information about the size and cost of that change. This is in large part because each stage implies a bigger sample size (fancy statistics-speak for “how many people interacted with the intervention”), a more durable design and process, a greater number of your organization’s people involved, and a greater likelihood that the intervention is going to become a permanent part of their standard operating procedure.

  • Pilots are tightly scoped interventions that we expect not to work (remember, we have to explicitly prove efficacy as a defense against confirmation bias), so we use small populations, focus on speed to market, and do them in an operationally dirty way.

  • Speed and resource efficiency are also important here. Because we chose multiple interventions during intervention selection, we’ll likely be running three to five concurrent pilots at any given time. If those pilots are too operationally heavy, we’ll stall out, so we have to be constantly focused on finding the lowest-fidelity version of an intervention that will still result in behavior change.

  • Pilot validation is just like insight validation: qualitative and quantitative confirmation that you’re headed in the right direction. Because of the small N, it won’t be statistically significant, but that’s fine; you’re just trying to get enough of a positive/negative/null signal to make a decision about what comes next. Did some people get flu shots at our church clinic who wouldn’t have otherwise? Did more people who got our letters get flu shots than didn’t? Get a rough idea, triangulate it, and move on. Or if it is truly null and you have the instinct that it simply is about not having enough signal, you can always enlarge a pilot and run it again.

  • At this point in the IDP, we’ve done multiple rounds of validation, extending from our very earliest insights and pressure mapping all the way to not one but two different validations of the actual intervention in the real world. We also have an idea of how much effort the intervention will take to deploy. So taken together, after the test, we can make a fairly strong juice/squeeze statement that sounds something like this:14 We are [confidence] that [intervention] will [direction] [behavior] (as measured by [data]). Scaling this requires [effort] and will result in [change]. Aren’t Mad Libs fun? Let’s do the decoding. Confidence = based on p-value but phrased colloquially Intervention = what the intervention is Direction = whether it increases or decreases the behavior Behavior = the measurable activity you established in your behavior statement Data = how you quantify that your population is doing the behavior Effort = the resources required to scale Change = based on p-value but phrased colloquially For our flu-shot letters, that might look something like: We are very confident that sending personalized flu shot letters based on member health motivations will increase the rate of getting a flu shot (as measured by flu shot claims). Scaling this requires about ten hours and $5,500 and will result in about five hundred additional flu shots.

  • Once you have insight into the cognitive preferences of your population, you may need to split the population into two or more as you drill down into the different places where people are willing to spend their energy. But this is where specificity and process help us. One of the values of the IDP is that it breaks down our products and services into smaller interventions that can be individually applied according to population. The difference between automation and curation is actually just one of interface: how many cognitive resources are people willing to spend on product choice? Everything else that makes a complete automatic or curatorial system is the same. Imagine we take over Blue Apron tomorrow. We know that it is currently configured for people who want to spend their attention on the act of cooking, not on recipes or ingredients. We also know there is a larger population that wants to spend their resources on just the eating. Should we ditch our current population in order to pursue the larger new one? Maybe, but that’s a false dichotomy. We can simply spin up another brand, Green Spatula, with recipes that are easier to cook. Same ingredients, same outcome, just the slightly less tasty version that doesn’t require you to have a vacuum sealer. Ninety-nine percent of the hard part of operating Blue Apron is the interventions and systems that allow for a reduction in cognitive spend around ingredients and recipes: the acquisition in bulk, the reduction to smaller quantities and combination into kits, and the timely delivery of those kits before things spoil. Getting that right was hard, as anyone at the company will tell you, and spinning up a second kind of kit for people who want an easier option is a cakewalk by comparison. A shockingly large number of companies can expand their market dramatically with relatively small interventions built on the same infrastructure. But we can do so only when we are specific in finding insights around where our populations want to spend their resources and basing interventions on those insights.

  • As a social psychologist, I make pretty much everything into a two-by-two matrix that yields four cells, and the uniqueness/belonging problem is no exception: let’s call the aspects stable and unstable, liker and disliker. They exist along a spectrum, but for convenience I’m going to talk about them as if they were discrete categories.

  • My preference for Johnny Cash is part of my snowflake, the part of me that is unique and special. And as a consequence, I’m a stable liker—my feelings about Johnny Cash are relatively impervious to what others think about him or his visibility in common culture. I didn’t like him more when Walk the Line racked up Oscar nominations or less when he made a commercial for Taco Bell. (If you’re wondering why that might make me like him less, find it on YouTube. “Where else can you get so many choices for just a little Cash?” is just . . . oof.) There are also stable dislikers. They’re wrong, but they exist—they’ve engaged meaningfully with Cash’s music and rejected it. Maybe they’re purists and feel like he was derivative of Son House and Robert Johnson and others. Maybe they just don’t like folk music. Whatever the reason, liking Johnny Cash doesn’t square with their identity and so they simply don’t, in a permanent way. This is part of their snowflake, core to their identity and unmodified by popular opinion, just like my preference for Cash. Then you have unstable likers of Johnny Cash. He wasn’t a part of their identity at all until Walk the Line came out, and suddenly they buy the American recordings and declare themselves fans. Just as they did with Ray Charles before him and Edith Piaf after, as their big-budget biopics came and went. Unstable likers are honoring the blizzard, trying to be a part of something that other people are also paying attention to because that’s the part of their identity that needs affirmation just then. It might be galling to some stable likers, who see themselves as the real fans, but it is just the other side of the coin and we all do it in different domains; seeking out a sense of belonging is important. And of course there are unstable dislikers (otherwise known as hipsters). They don’t care about Johnny Cash, they don’t care about Johnny Cash, they don’t care about Johnny Cash . . . and then boom! Walk. The. Line. Heavily edited Hollywood bullshit! He’s a sellout! Fuck Johnny Cash! Too bad he was already dead and so by definition couldn’t have sold out. But it doesn’t matter, because just like unstable likers, they’re in it for the belonging and the group and the shared hatred, until they find someone better to hate. And while they may feel easy to denigrate, unstable dislikers actually drive a lot of important creation: what is garage but a rejection of punk, punk but a rejection of rock, rock but a rejection of gospel? And that’s identity in a nutshell; anything that can turn gospel into garage rock in only a few leaps is worth paying attention to.

  • It is important to understand that we are all a part of each group, depending on the subject. We invest in a variety of identity signals because it makes it easier to balance both needs; a diversity of stable and unstable likes and dislikes ensures that we have the flexibility to adapt to environments and our own changing identity but also benefit from the permanence of identity anchors in a changing world. For example, I’m a stable liker of Johnny Cash and can come back to his music through thick and thin, whenever I need to honor my uniqueness. But I’m an unstable fan of most authors I’ve enjoyed; I’m currently hooked on Richard Kadrey (who I hope appreciates all the cursing in this book). But to be honest, I probably won’t remember his name in a year, given the rate at which I consume fiction and my tendency to ignore who actually wrote it.

  • So how do we find the stable and unstable likers and dislikers for our behavior of interest (or introspect about our own preferences; it’s okay, you’re not a narcissist, I promise)? After all, if we are going to start mapping the pressures associated with identities and creating interventions against them, we have to be able to target who is actually in that population. Fortunately, most of us can intuitively imagine ways to do this identification, simply because of the point I made earlier: we tend to spend an awful lot of our time, energy, and money on identity expression.

  • One of the signs of stability is a deep engagement, often with references to a personal connection to the behavior. It’s equally true for stable dislikes. I’m a stable disliker of thin-crust pizza, and you better believe that I could explain why, chapter and verse (thick crust, BBQ chicken with lots of onions, just to be clear). Two quick tests I often use are the TED test (could this person give an unprompted TED talk about this topic?) and the beer test (would this person be willing to go out for a beer and talk about nothing but this topic?).

  • People talking about unstable preferences just sound different. They tend to trail off fairly quickly when it comes to discussion, because their favorite is the group favorite, their stories are meant for quick group affiliation (I swear I’m one of you!), and they are unlikely to relate the subject to anything deeply personal to them. I like the writer Richard Kadrey because he swears and reflects on the complexity and toxicity of modern masculinity while mixing in demons and jokes and trips to Hell. But that’s about as far as it goes; I could probably fake a TED talk about Kadrey, but anyone from the stable camp would laugh at me.

  • Another easy trick is to look at self-signaling and social-signaling. If identity is the completion of the sentence “I’m the kind of person who . . . ,” self- versus social- is all about who is supposed to be reading that sentence. Almost all behaviors are targeted at both but usually lean heavily in one direction. Someone listening to Johnny Cash through their headphones is likely more of a stable liker, while pumping it to the whole office is a social, and thus more unstable, action.

  • If you want to find the stable, special, snowflake parts of a person, consider what they do when they’re alone. Yet another reason that triangulation between quantitative and qualitative validation is so important; qualitative researchers are good at differentiating between public and private behaviors, while quantitative researchers can use data from unfettered access to our unobserved moments. You can see the differences in the public sphere as well. People with stable preferences talk to you; people with unstable preferences talk about you (taken personally, this tip could save you years of therapy). Remember what identity needs are being honored in each case: stable is more about yourself and unstable is about the group. So who are stable people talking to? Themselves (ba-dum ching!). Or, as is more likely, directly to the artist, brand, etc. that they are engaged with. Unstable people, by contrast, are talking to one another. So while they may @ username, it generally isn’t in the first position precisely because they want everyone to see how much they like or dislike that subject. They need that message to show up in their followers’ feeds, because how else can they get responses? They’re actively recruiting for their interest, and for that they need a megaphone and some tribal signals of affiliation. I can’t count all the people I’ve told about Richard Kadrey in the last month or so, but I’ve had precisely zero conversations about Johnny Cash (other than with my editor, who remains unconvinced, though I’m determined to wear her down).

  • Each of the four groups could be thought of as a different population that you build different behavioral statements around, with different outcome behaviors, pressures, and interventions. In general, we don’t want the same behavior from each of the groups (and are unlikely to get it even if we do), so by starting to segment, we can flow with the tide of their behavior rather than against it. For each population, you can ask yourself, “What do I want from them?” and “What do they want from me?” and use those two questions to guide your process.

  • Let’s take Microsoft and stable likers. What does Microsoft want from stable likers? To buy, and keep buying, its products. But that’s true of everyone. What do stable likers have that others don’t? A deep knowledge about the Microsoft experience and the opinions to match. If you’re trying to create long-term behavior change, there is nobody quite like a stable liker to help you find novel insights, map pressures you didn’t even consider, and pilot your interventions. That’s why Microsoft has the Insiders programs. Windows Insider, for example, allows stable likers of Windows to engage directly with Microsoft as a brand. They get prerelease versions of the operating system, report bugs, and communicate directly with the Windows engineering team. More than sixteen million people have participated, and they generate petabytes (aka lots) of data every day. Try to imagine the equivalent number of quality-assurance testers you would need to employ to get the same coverage and you can see how “use a beta version of Windows” might be a very valuable outcome behavior, given that they’re working for free. And the best part? This is what stable likers want! If you are a stable liker of Microsoft, talking directly to the engineers and being involved in the process of creating the thing you like is precisely the kind of access you want, because it operationalizes your knowledge and deep engagement with the product. Unstable likers would have wanted a picture with Johnny Cash, anything they can brag about on social media and show off to the world and recruit others around. I would have wanted to take a walk with the man and just talk. Stable dislikers can also be used to discover insights and pressures, because they also have deep engagement with a subject, even if it is negative in conclusion. One of the offshoots of confirmation bias and our tendency to focus on promoting pressures when increasing behavior is the tendency to talk to the people who are already positively engaged. But the people who aren’t doing something (in a considered way) can be just as valuable for understanding behaviors. After all, who better than a stable disliker to identify the inhibiting pressures that tend to be our blind spot? And odd as it may seem, stable dislikers also want direct engagement. It is easy to get blinded by their dislike, but remember: they have chosen to form a stable identity around this subject. I am a stable disliker of a whole cabal of smart conservatives, but I’d still expend resources for the chance to take a walk with one and have a thorough conversation about where and why we differ. It would affirm my dislike and what makes me me at some core level, and that’s a perfectly fine motivation when what you want is perspective. Unstable likers have a unique behavior that we also create interventions around: because they are trying to build their tribe, they are natural-born recruiters. Stable likers aren’t as useful here, because they aren’t as concerned with what others think. But unstable likers love referral programs, buttons for social sharing, factoid-style content, and anything that allows them to actively share their temporary love of whatever. It can’t be deep; that’s a turnoff, as it highlights their unstable nature, and remember, nobody wants to be confronted with the reality that their preferences are impermanent. But as long as it supports their sense of belonging, unstable likers are all in.

  • Be warned! You have to be cautious of people with unstable preferences (once again, taken personally, more therapy avoided). Because their primary drive is belonging, they’re much more sensitive to the herd than their stable counterparts. Which means that if whatever group they want to be a part of happens to be made up of dislikers, they’ll drop you like a hot potato. As pointed out previously, this instability is actually good for our society as a whole, but if you were heavily invested in punk when everyone transitioned to garage, you could find yourself holding a lot of studded collars that you just couldn’t sell. Double down on continuous monitoring for any unstable intervention. But what about unstable dislikers? After all, they are unlikely to buy in to your primary behavioral statement, and yet you can’t ignore them, because they are out there vocally recruiting an army of haters to join them in disliking you. In reality, this is mostly about the other frame of reference that I mentioned early on in the book; sometimes we want to eliminate behaviors, which means strengthening inhibiting pressures and weakening promoting pressures. If the dominant promoting pressure of unstable likers is finding a group that embraces them, then one of the most effective ways to silence them is to show them that their behavior is unacceptable to the vast majority of the population. Which is why I love the Target social media team. Not the real one, although I’m sure they’re lovely as well, but “Ask ForHelp,” a fake Facebook account created by Mike Melgaard in 2015. Target had just announced a new policy to no longer segregate the toy aisles by gender, a move broadly met with acclaim but that also brought out unstable dislikers who just had to let the world know how much they loved gender norms. Because who doesn’t love a bunch of angry conservatives looking to find a sense of belonging and people who will agree with them? Good times.

  • After uploading the Target logo as his icon, Melgaard used the fake account to respond to the unstable dislikers on Target’s Facebook page as if he were the brand’s customer service team. Except instead of pandering to them, he mocked them, effectively shutting out any sense of community that might have grown around the pro-gender-norm commenters. His humorous responses got hundreds of likes before Target managed to have them taken down, while the rantings of the unstable likers typically received single digits.

  • The IDP is formulated to look at a behavior in isolation. This is deliberate, because one of the quickest ways to end up launching no interventions at all is to try to take in the entire scope of interrelated behaviors that make up a population. Competing behaviors is a twist based on a simple truth from our understanding of cognitive attention: at some level, everything competes with everything else. You can’t smoke and chew gum or take an Uber and watch Netflix on your couch at the same time. If you lower the price of your bottled water, people are going to drink less soda (why else do you think the soda companies own the water companies and keep the prices artificially high?). Every time we change a pressure or behavior, it affects every other pressure and behavior in some small way.

  • The fundamental mental backflip you have to make is simply that competing behaviors now require at least two sets of arrows instead of one. Normally, in a competing-pressures model, we increase promoting pressures and reduce inhibiting pressures if we want more of a behavior. But there is a second path. If we acknowledge that other behaviors compete, to greater or lesser degrees, with our behavioral goal, then it is fundamentally true that reducing alternative behaviors is part of increasing our outcome behavior. And thus trying to eliminate the alternative behaviors is a viable strategy. This can be dangerous, as you have to be careful about inadvertently reducing an overall pattern of behaviors. In startups, we often say, “A rising tide floats all ships,” meaning what is good for the industry is likely good for every startup in the industry, because even if your market share falls, the size of the overall market increases. Conversely, if you try to eliminate alternative behaviors that are too close to your behavior, you may capture more market share but shrink the overall market such that you find yourself with less than you had in the first place. Fortunately, this is where the IDP can help. When we start looking at a competing pressure, we essentially run an entirely new IDP focused on eliminating that behavior. But when we measure the pilots, instead of measuring the alternative behavior, we measure our true behavior of interest. Because we don’t actually care if water sales go up, so long as soda sales go down. Theoretically, we could repeat this process infinitely: one IDP for our outcome behavior, then an infinite number of IDPs addressing the alternative behaviors. But there are obviously diminishing marginal returns as we move further and further from likely alternative behaviors. If I weren’t writing this book, I’d be sleeping, so that’s a reasonable behavior to address. But the chances that instead I’d be skydiving are low, so an IDP there would be a waste. Picking the right alternative behavior is a little bit of a Goldilocks move: close enough that it is a viable alternative, far enough away that reducing it doesn’t inadvertently tank your outcome behavior. To avoid that conundrum, we don’t always have to try to extinguish the alternative behavior; instead, we can co-opt it. My favorite example is the former war between Uber and Netflix, and the reason I love it is that it was a war that nobody realized they were in. There was no TechCrunch article, no passive-aggressive digs at each other onstage, no drama of any kind. They may not even have realized they were fighting. And yet war it was. What did Uber want you to do on a Friday night? Go out (and preferably get drunk, so you couldn’t drive home). What did Netflix want you to do on a Friday night? Stay in. Those are mutually exclusive behaviors, and so a battle was required. This is one of the hidden beauties of the behavioral statement: it can help you find novel competitors and potential partners, simply by seeing whose limitations, motivations, and pressures align and collide with your own. Uber and Netflix resolved this through product development. Uber began delivering food (so it was fine if you wanted to stay in) and Netflix started doing mobile streaming (so it was fine if you went out, because you would watch Netflix in the back of an Uber). If they had really wanted to resolve the conflict, they could have given you free Netflix streaming while you were in the back of an Uber, maybe with zero rating of data usage brought to you by T-Mobile. These synergistic bundles are potential alternatives to running an IDP war against your competitors. Either way, these are big company tactics; if you’re small, stay focused on running the IDP for your own behavioral goals. Creating competition where it doesn’t already naturally exist is a waste of resources, and the whole point of starting at the end is getting more for less by focusing on why we do what we do. This is a right time, right place strategy.

  • At the most basic level, eliminating behavior uses the same IDP, except you are increasing inhibiting pressures and reducing promoting pressures instead of the reverse. And as with efforts to create more of a behavior, our pressure mapping will have that same predictable flaw: because we tend to focus on inhibiting pressures when we think about eliminating a behavior (punishments!), there is untapped upside in removing the promoting pressures that cause the behavior in the first place.

  • The basic problem is this: if we eliminate a behavior without replacing it with something else, the motivation goes unmet and people end up finding something else to fill in that behavioral gap. And often that replacement is quite a bit worse. Just ask Cady from Mean Girls; sure, she stopped Regina, but because she didn’t initially provide a pathway to honoring the teenage need to express identity, even worse behaviors sprang up. Hence the whole “break a crown and call out how we are all special snowflakes in a blizzard” ending; she had to eliminate the motivation to be a “mean girl.” Nature abhors a vacuum. Let’s take a non–Lindsay Lohan example: smoking. There is no public health victory quite so stirring as the reduction in smoking, from a peak in the midsixties, when about half of adult Americans smoked like chimneys, to its current rate of around one in ten and declining every year as older smokers die off. Cigarettes were ubiquitous, psychologically and biologically addictive, and one of the single most heavily marketed products of all time . . . and now they aren’t, all in about fifty years. Why? In accordance with our bias, we started with inhibiting pressures. Since death is pretty much the best inhibiting pressure you can get, we put giant fuck-off warnings on the packages. We started picking off the places you could smoke, one by one, turning smoking from a public behavior into a private one. We put in place massive taxes (more than half of the price of a modern pack of cigarettes is taxes) and aggressive laws about how cigarettes are sold, including strict prohibitions against any resale that would threaten those taxes. And this barrage of new inhibiting pressures worked, to an extent. But the reason we are where we are today, with a steadily declining smoking rate, is because of the more recent attacks on the promoting pressures. The most potent, of course, was the perception of smoking as cool. Think James Dean and the Marlboro Man. That’s why most smokers start in their teens, when the need for both uniqueness and belonging are at their peak. So that is where the fight began. If smoking was supposed to make you sultry and gorgeous like Marilyn Monroe, we’ll run ads of what smokers actually look like later in life. But because later is later and we want to target teens, we’ll threaten it right here and now. One of my favorite antismoking ads has no words. We see an attractive woman from behind. An attractive man notices her and begins to move in her direction, as if to ask her out. She turns and in profile is revealed to be smoking, causing him to veer away. The implication is clear: if you smoke, you won’t get laid. And for teens (and absolutely everyone else), getting laid is a powerful promoting pressure. Reduce it and presto, you get less smoking. And the cigarette companies can’t rebut the ad because we’ve banned their advertising. In movies, magazines, TV, billboards, radio, you name it—in America, you can advertise almost anything except cigarettes. Take away Big Tobacco’s ability to use the Marlboro Man or glamorize Virginia Slims as the latest weight-loss technique, and you weaken the behavior those promoting pressures create. Or, if you want to go even more ridiculous, think about cartoon characters. In the lawsuit that eventually ended Joe Camel, the plaintiff claimed that in the four years following his reintroduction, Camel cigarette sales to teens went from $6 million to $476 million. Around the same time, a Journal of the American Medical Association study found that six-year-olds were about as familiar with Joe Camel as with Mickey Mouse.20 Killing Joe Camel saved lives. Win one for a focus on behavior! Except now we have a problem: no more cigarettes, but the population (teens) still has the motivation (to look cool). They seek something that has ritual and routine, that honors their uniqueness and yet fosters belonging. Maybe something that you can have a favorite flavor of, that has endless accessories, that you can trade with your friends, that lets you ask that potential bae for a hit? We killed cigarettes but left a massive, gaping hole begging to be filled. Which might be why Juul (the largest e-cigarette company) just took a $13 billion investment from Altria (the largest cigarette company); everything about the behavioral statement, especially the motivation, stayed the same except that the behavior was e-cigarettes, so the synergy was natural. Thus the need for a replacement when focusing on eliminating a behavior. In ending smoking, we focused entirely on “stop smoking” and forgot that intrinsic need to give a pathway honoring the motivation, a “start something,” giving birth to the marginal improvement that is e-cigarettes.

  • Because ultimately, as much as we focus on behaviors, motivations are what matter. If we wanted to truly end smoking, we had to replace that behavior with something that helped teens continue to look cool and have a conversation starter (half of smoking is the social ritual of bumming a cigarette or a light from a stranger; the other half is using “Want to go have a smoke with me?” as a way of getting someone alone) before something else did.

  • I’ll end on a quick counterexample, just to show you what happens when you do it right. A few years ago, I was approached by a team of physicians working on the problem of heart disease in Africa. To combat a bland diet (motivation), many Africans had begun to add a large amount of sodium to their food (behavior). So we went to war on oversalting, with a variety of interventions. But we didn’t simply say, “Go back to bland food”; instead we introduced an array of non-salt spices that could be used to liven things up. Because ultimately, what they cared about was having food they actually wanted to eat; honor the motivation and you can eliminate salt, replace it with spice, and not worry about oil and sugar slipping in to fill the gap.