Start at the End: How to Build Products that Create Change
Authors: Matt Wallaert Tags: behavioral science, product management, innovation, psychology Publication Year: 2019
Overview
I wrote this book because the way most companies build products is fundamentally broken. It’s a relic of a ‘Mad Men’ era where people with privilege and power throw ideas around until one sounds sexy, build it, and then spend billions on advertising to convince people they should want it. This is an outrageous waste of time, money, and human potential. My central argument is simple: if you want to create change, you must start with the change you want to create. All products, services, and policies are interventions designed to alter human behavior, so we need to stop pretending otherwise. This book provides a rigorous, science-based framework called the [[Intervention Design Process (IDP)]] that flips the script. Instead of starting with a solution, you start by defining a specific, measurable behavioral outcome. Then, you work backward to understand the psychological forces at play—the ‘promoting pressures’ that make a behavior more likely and the ‘inhibiting pressures’ that make it less likely. The IDP is a step-by-step guide for product managers, designers, entrepreneurs, and anyone who wants to build things that matter. It’s for people who are tired of fetishizing process over outcomes and want a practical, repeatable way to apply behavioral science. In a world increasingly shaped by technology and AI, where the ethics of influence are paramount, this is not just a better way to build products—it’s a necessary one. It’s about replacing gut feelings with a systematic process of insight, validation, piloting, and testing to create interventions that genuinely work. This is my attempt at guerrilla warfare against waste and mediocrity, arming you with a process to build products that create real, meaningful change.
Book Distillation
1. The Intervention Design Process
The entire process for building products that change behavior starts with a potential insight—an observation about the gap between the world as it is and a better, counterfactual world. This insight is then validated and articulated in a precise behavioral statement. From there, the competing pressures are mapped, interventions are designed to modify those pressures, and an ethical check is performed. Finally, the chosen interventions are piloted, tested, and, if successful, scaled with continuous monitoring. This is the [[Intervention Design Process (IDP)]], the core framework for putting behavior at the center of creation.
Key Quote/Concept:
The Intervention Design Process (IDP): A systematic, multi-step framework for creating behavior change, moving from insight -> validation -> behavioral statement -> pressure mapping -> intervention design -> ethical check -> pilot -> test -> scale.
2. Potential Insights and Insight Validation
A potential insight is the recognition of a gap between two universes: the one we live in and a more optimal one. Insights are the raw material for behavior change and come in four main types: quantitative (from data patterns), qualitative (from observation), apocryphal (common knowledge in an organization), and external (from outside research). No insight should be taken as truth until it’s validated using [[convergent validity]]—triangulating evidence from multiple, diverse sources to fight confirmation bias.
Key Quote/Concept:
Four Types of Insights: Insights can be quantitative, qualitative, apocryphal, or external. The key is not the source but the validation; always seek convergent validity from multiple sources before proceeding.
3. Behavioral Statement
To put behavior at the center of the creation process, you must clearly articulate the desired outcome. A behavioral statement is a precise, falsifiable, and non-negotiable definition of the world you are trying to create. It forces clarity and alignment, preventing teams from getting fixated on features instead of the end goal. This statement acts as the North Star for the entire IDP.
Key Quote/Concept:
The Behavioral Statement Formula: ‘When [population] wants to [motivation], and they [limitations], they will [behavior] (as measured by [data]).’ This structure forces you to define every critical variable of the change you want to see.
4. Pressure Mapping and Pressure Validation
Behavior is the net result of competing forces. [[Promoting pressures]] make a behavior more likely, while [[inhibiting pressures]] make it less likely. To change behavior, you must alter this balance. There is a natural bias to focus only on adding promoting pressures (like rewards or new features), but there is enormous, often untapped, power in identifying and removing inhibiting pressures (like friction, cost, or ambiguity).
Key Quote/Concept:
Competing Pressures: Visualize behavior as a balance between an up arrow (promoting pressures) and a down arrow (inhibiting pressures). Effective intervention design requires mapping and validating the forces on both sides of the equation.
5. Intervention Design and Intervention Selection
Once pressures are mapped, the goal is to generate a high volume of potential interventions that directly address them. The focus should be on creating a diverse set of options, not just the most creative or obvious ones. From this large pool, a small number of interventions are selected for piloting. The selection should aim for [[optimum distinctiveness]]—a range of different approaches that cover the problem space without too much overlap, maximizing what can be learned.
Key Quote/Concept:
Optimum Distinctiveness: When selecting interventions to pilot, don’t just pick variations of the same idea. Choose a range of distinct approaches (like strawberry, marmalade, and kiwi jam) to maximize learning about what works.
6. Ethical Check
Changing behavior carries an inherent ethical responsibility. Before piloting any intervention, it must be checked against a clear ethical framework. This involves ensuring the behavior aligns with a pre-existing motivation in the population (closing an [[intention-action gap]], not creating a new goal). The benefits of the behavior change must not be outweighed by the costs to an alternative motivation, and you must be willing to be transparent about your goals and methods.
Key Quote/Concept:
The Ethical Rule: An intervention is unethical if (1) the outcome behavior is not the result of any of the population’s existing motivations, (2) its benefit doesn’t outweigh the cost to an alternative motivation, or (3) you are unwilling to publicly describe and take responsibility for it.
7. Pilot and Pilot Validation, Test and Test Validation, Scale Decision and Continuous Measurement
Implementation is a staged process. A [[pilot]] is a small-scale, ‘operationally dirty’ experiment designed to get a quick signal on whether an intervention works at all. A [[test]] is a larger, more robust trial to determine if the intervention is worth the effort to scale. The outcome of a test is a scale decision, articulated in a ‘juice/squeeze statement’ that weighs the expected impact against the required resources. Scaled interventions require continuous monitoring because their effectiveness can decay over time.
Key Quote/Concept:
Pilot, Test, Scale: This three-stage process de-risks implementation. Pilots are for signal (Does it work at all?). Tests are for efficiency (Is the juice worth the squeeze?). Scale is for impact, and requires continuous measurement.
8. The End of the Beginning
Part 1 provides the complete, basic framework of the IDP. Creating behavior change is messy, but having a systematic process makes it manageable. The IDP can be applied not just to external customers but also to internal organizational challenges. If your company culture is an obstacle, define the desired cultural behavior, map the pressures, and run interventions. The process is universal because people are people.
Key Quote/Concept:
Apply the IDP Internally: If your organization is resistant to this process, don’t give up. Treat the organizational resistance as a behavior change problem in itself and apply the Intervention Design Process to change your own company.
9. Priming, Moderation, and Mediation
[[Identity]] is the most powerful pressure influencing behavior. Instead of vague personas, think of identity as a hierarchy of roles, values, and behaviors. You can leverage identity with three techniques. [[Priming]] activates an existing, strong link between an identity and a behavior. [[Moderation]] strengthens or weakens an existing link. [[Mediation]] creates a new link between a behavior and a core value, connecting it to a pre-existing identity.
Key Quote/Concept:
Priming, Moderation, Mediation: These are three ways to use identity. Priming activates an identity (reminding a student they are a scholar before a test). Moderation changes the strength of an association (‘run like a girl’). Mediation creates a new association (linking computer use to the value of ‘caring’ for nurses).
10. Optimum Cognition
Cognitive attention is a finite resource, and the brain is a [[cognitive miser]], hardwired to conserve it. Every product and service competes for a slice of this dwindling pie. Effective design requires understanding where your users want to spend their cognitive energy and where they don’t. This leads to designing for automation (for things they don’t care about) or curation (for things they do), and for satisficing (finding ‘good enough’) or maximizing (finding the absolute best).
Key Quote/Concept:
Cognitive Miser: The brain is lazy and wants to conserve energy. Design for this reality by aligning the cognitive load of your product with the user’s desire to spend mental energy on that specific task.
11. Uniqueness and Belonging
Humans are defined by a fundamental paradox: the need to be a unique individual (the snowflake) and the need to be part of a group (the blizzard). This tension is a powerful driver of behavior. Interventions can cater to one or both needs, for example by offering customization (uniqueness) and then making that customization shareable (belonging). Understanding whether your audience is oriented more toward uniqueness or belonging can unlock powerful design strategies.
Key Quote/Concept:
The Snowflake-in-a-Blizzard Problem: People are constantly balancing the need to stand out with the need to fit in. This creates four identity types relative to any topic: stable likers, unstable likers, stable dislikers, and unstable dislikers, each requiring a different engagement strategy.
12. Special Factors of Inhibiting Pressures
Focusing on removing [[inhibiting pressures]] is disproportionately powerful. Because of our natural bias toward adding promoting pressures, inhibiting pressures are often ignored and full of low-hanging fruit. They tend to be more universal, durable, and predictable than promoting pressures. Most importantly, due to [[prospect theory]], removing a negative (a loss) is psychologically more powerful than adding an equivalent positive (a gain).
Key Quote/Concept:
Prospect Theory in Product Design: ‘Equivalent losses hurt worse than equivalent gains feel good.’ This means that removing a $5 friction (an inhibiting pressure) is often more powerful than adding a $5 reward (a promoting pressure).
13. Competing Behaviors
No behavior exists in a vacuum; at some level, everything competes with everything else for time, money, and attention. Instead of only focusing on making your desired behavior more likely, you can also work to make a [[competing behavior]] less likely. For example, to increase savings, you can either make saving easier or make spending harder. This opens up a new strategic path for intervention.
Key Quote/Concept:
The War Nobody Sees: Behaviors compete. Netflix (staying in) and Uber (going out) were in a war for Friday night. They resolved it through product development: Uber started delivering food (supporting staying in) and Netflix went mobile (supporting going out).
14. Eliminating and Replacing Behavior
When you eliminate a behavior, the underlying motivation that drove it remains. If you don’t provide a positive [[replacement]] that honors that motivation, people will find their own, and it might be worse than the original behavior. The successful campaign to reduce teen smoking worked by attacking the promoting pressure (‘coolness’) but left the motivation unmet, creating a vacuum that e-cigarettes (like Juul) were perfectly designed to fill.
Key Quote/Concept:
Nature Abhors a Vacuum: To successfully eliminate a negative behavior, you must also introduce a positive replacement that satisfies the original motivation. Otherwise, a new, potentially worse behavior will fill the void.
15. Mini Case Studies
The parable of the Good Samaritan illustrates a core principle of behavior change. While we tend to believe helping is driven by internal promoting pressures like character and morality, experiments show it is often dictated by inhibiting pressures. Seminary students on their way to give a talk about the Good Samaritan were far less likely to stop and help an injured person if they were told they were running late. The inhibiting pressure of ‘timeliness’ overwhelmed the powerful promoting pressure of their identity and immediate context.
Key Quote/Concept:
The Hurried Samaritan: A powerful inhibiting pressure (like being in a hurry) can easily overpower a strong promoting pressure (like a person’s core values). This demonstrates why understanding and addressing inhibiting pressures is critical.
Generated using Google GenAI
Essential Questions
1. Why is the traditional ‘Mad Men’ approach to product development broken, and what is the core principle of the Intervention Design Process (IDP) that replaces it?
I argue the ‘Mad Men’ model—where powerful people build sexy-sounding ideas and then use advertising to create demand—is fundamentally wasteful and ineffective. It fetishizes the solution and the process over the actual outcome. This approach starts with an intervention (the product) and hopes it causes a desired behavior. The core principle of my [[Intervention Design Process (IDP)]] is to flip this script entirely. You must ‘start at the end’ by defining a specific, measurable behavioral outcome first. This is articulated in a rigorous [[behavioral statement]]. Only after defining the desired change do you work backward to understand the existing psychological forces—the [[promoting pressures]] and [[inhibiting pressures]]—that govern the current behavior. The product, then, is not the goal; it is merely an intervention designed specifically to alter that balance of pressures and achieve the pre-defined behavioral outcome. This shifts the focus from building ‘cool stuff’ to building ‘stuff that works’ in a measurable, scientific, and repeatable way.
2. How does the concept of competing pressures reframe the problem of behavior change, and why is focusing on inhibiting pressures often more powerful?
Behavior is not the result of a single cause but the net product of competing forces. I visualize this as a balance: [[promoting pressures]] push a behavior to be more likely, while [[inhibiting pressures]] push it to be less likely. To create change, you must alter this balance. The traditional product mindset has a strong bias toward adding promoting pressures—new features, rewards, better marketing—because it feels creative and additive. However, this ignores the immense, often untapped, power on the other side of the equation. Focusing on removing [[inhibiting pressures]] is disproportionately powerful for several reasons. First, due to our collective blind spot, there is less competition and more low-hanging fruit. Second, as [[prospect theory]] demonstrates, removing a negative (a loss, like friction or cost) is psychologically more impactful than adding an equivalent positive (a gain). Finally, inhibiting pressures like cost and inconvenience tend to be more universal, durable, and predictable than promoting pressures like taste or desire, making interventions that target them more efficient and scalable.
3. What is a behavioral statement, and how does its specific structure serve as the North Star for the entire product development process?
A behavioral statement is a precise, non-negotiable, and falsifiable articulation of the change you want to create. It’s the cornerstone of the IDP because it forces absolute clarity on the outcome before any work on a solution begins. The structure is: ‘When [population] wants to [motivation], and they [limitations], they will [behavior] (as measured by [data]).’ Each component is critical. It defines who you’re targeting, their pre-existing motivation (which is key for ethical intervention), the limitations that are outside your control, the specific behavior you aim to produce, and the data that will prove you’ve succeeded. This structure prevents teams from falling in love with features or processes. Instead of a vague goal like ‘increase engagement,’ the team has a concrete, measurable objective. It serves as the ultimate arbiter for every decision: does this proposed intervention directly map to a pressure that will cause this specific behavior for this specific population? It aligns the entire organization around a single, clear definition of success.
4. What is the ethical framework for behavior change, and how does it distinguish between closing an intention-action gap and creating new motivations?
Changing behavior carries an immense ethical responsibility. My framework is designed to ensure we are helping people, not manipulating them. The core distinction is between closing an [[intention-action gap]] (helping someone do what they already want to do) and creating a new goal. The ethical rule has three parts: an intervention is unethical if (1) the outcome behavior is not the result of any of the population’s existing motivations, (2) its benefit doesn’t outweigh the cost to an alternative motivation, or (3) you are unwilling to publicly describe and take responsibility for it. The first point is crucial: we must anchor our work in a motivation the user already has. For example, helping someone who wants to save money actually save money is ethical. Creating a desire to buy a product they don’t need is not. This framework explicitly rejects designing for addiction or habit formation that circumvents a user’s original intentions. It forces us to be transparent and to weigh the holistic impact of our work, ensuring we are empowering users, not exploiting their psychological biases for our own gain.
Key Takeaways
1. Start with the Behavioral Outcome, Not the Solution
The most fundamental principle of the book is to invert the typical product development process. Instead of starting with an idea for a product or feature and then figuring out how to market it, you must start by defining a precise, measurable behavioral outcome. This is codified in the [[Behavioral Statement]]. This approach forces clarity, alignment, and discipline. It ensures that every subsequent step—from research to design to testing—is laser-focused on achieving that specific change. By falling in love with the problem (the gap between current and desired behavior) instead of a particular solution, teams are free to design the most effective intervention, rather than the one that sounded best in the initial brainstorm. This outcome-first orientation is the antidote to building products that nobody needs and wasting resources on features that don’t drive meaningful change.
Practical Application: An AI product engineer working on a new feature for a learning platform should not start with ‘Let’s build an AI-powered chatbot.’ Instead, they should start by defining a behavioral statement like: ‘When a student wants to master a difficult concept, and they are studying alone, they will use our platform to practice that concept for 15 minutes (as measured by active engagement time).’ This frames the entire project around a user’s goal and a measurable behavior, ensuring the AI chatbot (or any other solution) is built and judged solely on its ability to achieve that outcome.
2. Removing Inhibiting Pressures Is More Powerful Than Adding Promoting Pressures
Every behavior is governed by a balance of [[promoting pressures]] (reasons to do it) and [[inhibiting pressures]] (reasons not to). Product teams have a natural bias to focus on adding promoting pressures, like new features, rewards, or better aesthetics. However, the book argues that the greatest opportunities often lie in identifying and removing inhibiting pressures like friction, cost, ambiguity, or complexity. This is powerful for two reasons. First, due to [[prospect theory]], people feel the pain of a loss (like wasting time) more acutely than the pleasure of an equivalent gain (like a reward). Removing friction is therefore psychologically potent. Second, because this area is often ignored, it’s full of low-hanging fruit. Instead of trying to out-feature competitors, you can gain a significant advantage by making the desired behavior radically simpler or easier to perform.
Practical Application: An AI product engineer at an e-commerce company wants to increase the use of a new AI-powered recommendation engine. Instead of adding more promoting pressures (e.g., making the recommendations flashier), they should map the inhibiting pressures. Perhaps the recommendations appear too late in the checkout flow (friction), the language is confusing (ambiguity), or it slows down page load time (cost). By systematically removing these inhibitors, they can dramatically increase adoption without changing the core recommendation algorithm at all.
3. De-Risk Innovation with a Staged Process: Pilot, Test, Scale
The book advocates for a disciplined, staged approach to implementation that de-risks innovation and prevents large-scale failures. This process moves from a [[pilot]] to a [[test]] to a scale decision. A pilot is a small-scale, ‘operationally dirty’ experiment designed to get a quick, cheap signal on whether an intervention works at all. It prioritizes speed and learning over polish. If a pilot shows a positive signal, it moves to a test. A test is a larger, more robust trial designed to answer the question, ‘Is the juice worth the squeeze?’ It validates whether the intervention is efficient and effective enough to be worth the resources required to build it properly. Only after a successful test does a team make a scale decision, backed by a clear ‘juice/squeeze statement’ that quantifies the expected impact versus the cost. This systematic process replaces big, risky bets with a series of smaller, evidence-based decisions.
Practical Application: An AI product engineer has an idea for a complex new fraud detection model. Instead of spending six months building the full model, they should first run a pilot. This could be a simple heuristic model applied to 1% of traffic for one week, or even a manual review process that simulates the model’s logic (‘Wizard of Oz’ pilot). The goal is just to see if the concept has any signal. If it does, they can proceed to a test: building a more robust version of the model and running it on a larger user base to get statistically significant data on its accuracy and operational cost. Only then do they decide whether to scale it across the entire system.
Suggested Deep Dive
Chapter: Chapter 12: Special Factors of Inhibiting Pressures
Reason: This chapter is the tactical heart of the book’s counter-intuitive strategy. While the entire book emphasizes balancing pressures, this section explains why an asymmetrical focus on removing friction, cost, and ambiguity is often the most effective path to behavior change. It provides the theoretical backing (like [[prospect theory]]) and practical reasoning (they are more universal, durable, and ignored) that can help a product engineer win arguments and prioritize work that delivers outsized impact. For anyone building complex systems like AI, where user friction can be high, mastering the art of identifying and eliminating inhibiting pressures is a superpower.
Key Vignette
Bing in the Classroom: The IDP in Action
The process began with a potential insight at Microsoft: ‘Kids don’t search in school nearly as much as you’d think.’ Instead of jumping to a solution, my team first validated the insight with data (queries per student were less than one) and qualitative observation. The resulting behavioral statement focused on increasing student searches. When mapping the [[competing pressures]], we discovered the problem wasn’t a lack of student curiosity (a promoting pressure), but overwhelming inhibiting pressures from teachers, who worried about inappropriate content, advertising, and student privacy. Therefore, the intervention wasn’t a splashy ad campaign about curiosity, but a specialized version of Bing for schools with SafeSearch locked on, no ads, and reduced data collection, bundled with structured lesson plans. This targeted removal of inhibiting pressures led to a 40% increase in school searches and a successful, scaled product.
Memorable Quotes
Instead of starting at the end—a clearly described behavior that is the explicit goal of creation—our method of design has come to embrace the sexy sell and the need to sound good instead of be good.
— Page 13, Introduction
When [population] wants to [motivation], and they [limitations], they will [behavior] (as measured by [data]).
— Page 38, Chapter 3: Behavioral Statement
Good behavioral design uses science as a process but isn’t about producing pure truths, and sometimes currying favor with the marketing department isn’t such a bad thing. Repeat after me: we are outcome focused, we are outcome focused, we are outcome focused. Say it until you believe it.
— Page 25, Chapter 1: The Intervention Design Process
Nature abhors a vacuum: To successfully eliminate a negative behavior, you must also introduce a positive replacement that satisfies the original motivation. Otherwise, a new, potentially worse behavior will fill the void.
— Page 136, Chapter 14: Eliminating and Replacing Behavior
The temptation is to judge them harshly, but think back to your own behavior: have you ever felt in a hurry and snapped at a laggard child or gruffly pushed your way through a crowd, things you would never do in a more relaxed moment? Perceptions of lateness don’t create the promoting pressures… But as an inhibiting pressure, how late we feel in the moment is one of the strongest factors in modifying behavior.
— Page 141, Chapter 15: Mini Case Studies
Comparative Analysis
I wrote ‘Start at the End’ to be a practical field manual, which sets it apart from other foundational texts in behavioral science. While Daniel Kahneman’s ‘Thinking, Fast and Slow’ brilliantly explains the ‘why’ of human irrationality and Richard Thaler and Cass Sunstein’s ‘Nudge’ provides a powerful philosophy for choice architecture, my book provides the ‘how.’ It’s not a collection of interesting studies; it’s a step-by-step, repeatable framework—the [[Intervention Design Process (IDP)]]—for applying those insights to build actual products and policies. This makes it a direct complement to those works. It also stands in stark contrast to a book like Nir Eyal’s ‘Hooked,’ which focuses on building habit-forming products. I explicitly argue that designing for habit is often unethical because it can create motivations rather than serving existing ones. My ethical framework, centered on closing the [[intention-action gap]], provides a crucial guardrail for product builders, especially in the AI space, where the potential for unintentional (or intentional) manipulation is high. My work is less about theory and more about a disciplined, ethical, and effective process for creating measurable change.
Reflection
My goal with this book was to wage guerrilla warfare against waste. The waste of money, talent, and time that comes from building products based on gut feelings and post-hoc rationalizations. The [[Intervention Design Process (IDP)]] is my weapon. Its strength lies in its relentless focus on a single, measurable behavioral outcome, which provides clarity and a defense against corporate politics and shiny object syndrome. It forces scientific rigor—validation, testing, measurement—into a field often dominated by subjective creativity. The book’s weakness, if you can call it that, is its strident and informal voice. This was a conscious choice. The ‘professional’ tone of most business books often serves to uphold the very ‘Mad Men’ culture I’m trying to dismantle. I wanted to write something accessible and authentic, even if it’s not for everyone. In the broader context of AI development, this process is more critical than ever. AI provides an unprecedented ability to influence human behavior at scale. Without a rigorous, ethical, and outcome-focused framework like the IDP, AI product engineers risk building powerful systems that optimize for the wrong metrics, creating unintended negative consequences. This book is a plea to build with purpose, to measure what matters, and to always, always start at the end.
Flashcards
Card 1
Front: What is the core principle of the [[Intervention Design Process (IDP)]]?
Back: To ‘start at the end’ by defining a specific, measurable behavioral outcome first, and then working backward to design an intervention that achieves it.
Card 2
Front: What are the two types of [[competing pressures]] that determine behavior?
Back:
- [[Promoting pressures]]: Forces that make a behavior more likely (e.g., desire, rewards). 2. [[Inhibiting pressures]]: Forces that make a behavior less likely (e.g., friction, cost, ambiguity).
Card 3
Front: What is the formula for a [[Behavioral Statement]]?
Back: ‘When [population] wants to [motivation], and they [limitations], they will [behavior] (as measured by [data]).’
Card 4
Front: What is the key ethical distinction for a behavior change intervention?
Back: Whether it closes an existing [[intention-action gap]] (helping people do what they already want) versus unethically creating a new motivation or goal.
Card 5
Front: What is the difference between a [[pilot]] and a [[test]] in the IDP?
Back: A pilot is a small, fast, ‘operationally dirty’ experiment to get a signal (Does it work at all?). A test is a larger, more robust trial to determine efficiency (Is the juice worth the squeeze?).
Card 6
Front: According to [[prospect theory]], why is removing a $10 fee often more powerful than giving a $10 discount?
Back: Because ‘equivalent losses hurt worse than equivalent gains feel good.’ Removing the pain of the fee (a loss) has a greater psychological impact than adding the pleasure of the discount (a gain).
Card 7
Front: What is the ‘cognitive miser’ principle?
Back: The brain is hardwired to conserve mental energy. This means product design must align the cognitive load of a task with the user’s desire to spend mental energy on it, designing for automation or curation accordingly.
Card 8
Front: What is the ‘Nature Abhors a Vacuum’ principle in behavior elimination?
Back: When you eliminate a negative behavior, you must provide a positive [[replacement]] that satisfies the original motivation, or a new, potentially worse behavior will fill the void.
Generated using Google GenAI