Autonomy, competence, relatedness: the three reasons people stay (or leave) your product

Every product team I have worked with has had a theory about why users leave. The onboarding is too long. The value proposition is unclear. There is no habit loop. Some of those are true, but they are symptoms. The underlying question is: does using this product satisfy a basic psychological need, or does it tax one? Edward Deci and Richard Ryan's self-determination theory (SDT) gives a framework for that question that has been tested across cultures, age groups, and domains for over forty years.

The three needs

Autonomy: "I chose this"

Autonomy is not about having infinite options—that is the paradox of choice, which is a different problem (Schwartz, 2004). Autonomy is about feeling that your actions originate from your own values and intentions, not from external pressure. In product terms, this means letting users customise without requiring it, offering opt-in over opt-out defaults, and never making someone feel trapped. Dark patterns violate autonomy directly, and the damage is not just ethical—it undermines the intrinsic motivation that keeps people coming back without a push notification.

Barry Schwartz's work is worth pairing with SDT here. His research (Schwartz, 2004; Iyengar & Lepper, 2000) found that more options can decrease satisfaction and increase regret—particularly for "maximisers" who feel compelled to evaluate every possibility. So autonomy-supportive design is not "give them every setting." It is "give them meaningful choices with sensible defaults." Respect their agency without burdening their attention.

Competence: "I can do this"

Competence is the need to feel effective—to master challenges and see results from your effort. Csikszentmihalyi's flow research (1990) overlaps here: flow happens when challenge and skill are in balance, goals are clear, and feedback is immediate. Too easy and you get boredom. Too hard and you get anxiety. The sweet spot is where the product stretches you just enough that success feels earned.

This is where the goal gradient effect matters. Clark Hull (1932) first observed that rats run faster as they approach a reward. Kivetz, Urminsky, and Zheng (2006) demonstrated the same pattern in humans with loyalty cards: customers who received a card pre-stamped with two purchases (of twelve required) completed the card faster than those who started from zero on a ten-stamp card—even though both needed ten purchases. Perceived progress drives effort. In product terms: progress bars, step indicators, and visible streaks work because they tap competence. They say "you are getting somewhere."

Norton, Mochon, and Ariely's IKEA effect research (2012) adds another layer. When people invest effort into building or customising something, they value the result more—sometimes irrationally so. That means onboarding steps where users actively configure a dashboard or choose preferences are not wasted friction. They are investment that creates psychological ownership. The key is that the task must be completable. Incomplete or failed assembly kills the effect.

Relatedness: "I belong here"

Relatedness is the need to feel connected and understood. It does not require a social feature. A product satisfies relatedness when its tone feels human, when its support feels personal, when using it connects you to a community or a shared purpose—even implicitly. Cialdini's unity principle (2021) is related: shared identity increases cooperation and trust. If your product makes users feel like insiders rather than transaction numbers, you are designing for relatedness.

Deci's early experiments (1971) showed that external rewards can actually undermine intrinsic motivation if they feel controlling. The product equivalent: a push notification that says "You have not logged in for three days!" does not support relatedness. It signals surveillance. A message that says "Your team finished the sprint—here is what shipped" supports both competence and relatedness without coercion.

The Fogg angle: ability is usually the bottleneck

BJ Fogg's behavior model (B = MAP: Motivation × Ability × Prompt) is worth layering on top. Fogg's insight—validated through Stanford's Behavior Design Lab—is that most products over-invest in motivation and under-invest in ability. You do not need a more compelling reason to act if the action is effortless enough. Reducing form fields from six to two, pre-filling data, collapsing a three-screen flow into one: these are ability interventions, and they often outperform motivational copy by a wide margin.

Where SDT and Fogg intersect is that supporting autonomy and competence is itself an ability intervention. A user who feels capable and in control perceives the task as easier. Perceived difficulty is subjective, and it is shaped by the emotional context SDT describes.

Open loops and the Zeigarnik effect

Bluma Zeigarnik (1927) found that people remember incomplete tasks roughly ninety percent better than completed ones. The mechanism is cognitive tension: an unresolved task occupies working memory, demanding closure. Product designers use this constantly—progress bars at 65%, partially completed profiles, unread badge counts. It works. But it also has a limit. If every surface of your product screams "finish me," you have traded engagement for anxiety. Ethical application of Zeigarnik means creating one or two purposeful open loops, not twenty.

What I actually changed

• Onboarding now includes one meaningful customisation step (IKEA effect) with a visible progress bar (goal gradient). The step must be completable in under sixty seconds.
• Default settings are generous—users can adjust but do not have to. Autonomy without the paradox-of-choice tax.
• Error messages name the problem and the next step. No dead ends. Competence means the user can always recover.
• Notification copy is descriptive, not commanding. "Your report is ready" rather than "Don't forget to check your report!"
• Session-end screens restate what was accomplished, not what is left undone. Peak–end thinking meets competence.

None of this is revolutionary. What changed for me was having a framework that explains why these details compound. SDT, Fogg, Zeigarnik, flow: they are not competing theories. They are different lenses on the same question—does this product make someone feel capable, autonomous, and connected, or does it drain them? If the answer is drain, no amount of gamification will fix retention.