Which would you rather experience:
- A last-minute, surprise trip to Europe?
- Or a trip to Europe with a month's advance notice?
Most people pick the second option, and not only because it gives them time to clear their work calendar. Human nature enjoys anticipation. It adds energy to an experience before the experience even starts.
Neuroscience backs this up. The nucleus accumbens, a region tied to how we register pleasure, activates when a healthy brain experiences something it finds rewarding. For years, researchers assumed that activation happened at the moment of the reward itself. Stanford's Brian Knutson overturned that assumption: the spike happens during anticipation, before the reward ever arrives. A 2023 study in the Journal of Neuroscience found the same pattern specifically in operant tasks, where dopamine tracked the anticipation of an action's payoff rather than the payoff itself.
That single fact explains a lot of behavior that doesn't look pleasurable on its own but gets repeated anyway. If your brain lights up in anticipation, you'll keep doing the thing that comes right before the payoff, whether or not the payoff is guaranteed. That's the mechanism behind variable rewards, and it's still the backbone of how the most-used apps on your phone are built. This piece was first written in 2018, when that idea was still fairly new to product teams. A lot has happened since, and it's worth walking through what still holds, what doesn't, and where the line sits now between smart design and manipulation.
The Science Behind Variable Rewards
Long before Knutson's work, B.F. Skinner was already studying how reward schedules shape behavior. In his famous "Skinner Box" experiments from the 1930s, a hungry rat learned that pressing a lever produced food. Once it made that connection, it kept pressing.
Then Skinner changed the rules: instead of a reward every time, the lever paid out unpredictably. The rat pressed more, and it kept pressing even during stretches when the lever produced nothing at all. Skinner's conclusion still holds up: unpredictable rewards produce more persistent behavior than predictable ones, and that behavior is harder to break once it's established.
Three Types of Variable Rewards
Author and behavioral design consultant Nir Eyal, writing about the same mechanism for product teams, grouped variable rewards into three categories: rewards of the tribe, the hunt, and the self.
Rewards of the Tribe
People are wired to need each other. We look for acceptance and a sense of belonging, and platforms built around visible social approval tap directly into that.
Reddit is a clean example, and it's not a shrinking one. The platform hit 121.4 million daily active users in the last quarter of 2025, up 19% year over year. Karma and awards give contributors an unpredictable, public signal of whether their community values what they said, and that uncertainty is a big part of what keeps people posting.
Quora shows the opposite pull. It's lost roughly 28% of its visitors over the past year, a decline that tracks with people asking AI chatbots questions instead of posting them publicly. An AI chat can answer the question fine, and you'll eventually get what you need. But for a community-driven product, that was never really the reward. Knowing your question or comment sparked a real discussion and that other people engaged with it is a different kind of value than just getting an answer. That's real human interaction, and it's the one ingredient no AI chat can fake.
Rewards of the Hunt
Hunting is a primal behavior. Most of us aren't hunting for food anymore, but we're still wired to chase things that feel scarce, whether that's a deal at checkout or a winning hand of cards.
Casinos remain the textbook case. But the sharper modern example is the algorithmic feed. TikTok's For You page and Instagram Reels are built around an endless, unpredictable hunt for the next good video, and unlike a 2018 social feed, the algorithm is actively learning what kind of unpredictability keeps a given person scrolling the longest. That's the hunt reward with a machine adjusting the odds behind the scenes, in real time, per user.
Rewards of the Self
We like finishing things. Checking a box or watching a streak counter climb feels good on its own, apart from any social payoff at all.
Duolingo is the clearest current case study. In its second-quarter 2026 results, Duolingo reported 58.7 million daily active users and an 84% next-day return rate, driven in large part by its streak mechanic. When the company ran a "Streak Revival" campaign letting lapsed users restore their longest streak after finishing three lessons, 15.4 million learners took the offer. Nearly 8 million of them had no active streak at all when the campaign started. That's the reward of the self, running at a scale Skinner's rat never got to test.
How AI-Driven Algorithms Supercharged Variable Rewards
The biggest change since 2018 is what's doing the tuning. Skinner varied a lever by hand. A 2018 Facebook feed varied in roughly the same way for every user. A 2026 recommendation algorithm varies the reward individually, in real time, for each person, based on exactly what has and hasn't worked on that specific person before. That's a different order of magnitude. The mechanism Skinner documented with one rat and one lever is now running millions of personalized experiments a day, each one tuned to keep one specific person anticipating whatever comes next in their feed.
The Ethics and Regulation of Variable Rewards
Back in 2018, the only ethical warning in this article was a short note about privacy. That's not enough anymore. In the last few years, "is this manipulative?" has stopped being just a design question. It's now a public health question, and in some places, a legal one too.
In 2023, the U.S. Surgeon General published an official advisory on social media and youth mental health. It said there isn't enough evidence yet to call social media safe for kids and teens, and it urged companies, lawmakers, and parents to take that seriously. In 2024, psychologist Jonathan Haidt made a similar case to a much bigger audience in his book The Anxious Generation. His argument: handing kids phones and apps built on this same "you never know what you'll get" reward system is a major reason anxiety and depression have climbed so sharply among teenagers.
Regulators have started paying closer attention too, though it's a mixed picture so far. In 2024, the FTC and its international partners reviewed 642 subscription websites and apps and found that 76% used at least one "dark pattern," a design trick that pushes people toward a choice they wouldn't otherwise make, like burying the cancel button. The European Commission is also expected to propose a new Digital Fairness Act sometime before the end of 2026, aimed specifically at addictive app design. But it isn't a straight march toward stricter rules everywhere: in mid-2025, a U.S. appeals court struck down the FTC's "click-to-cancel" rule, which would have required companies to make canceling a subscription as easy as signing up for one. Even so, both the U.S. and the EU are generally headed toward more scrutiny of the kind of reward-driven design this article is about.
Designli Approach: Building Reward Loops That Earn Their Place
Every variable reward mechanic in this article works on the same premise: give people something worth anticipating. The ones that backfire skip the harder question, which is whether the product underneath the mechanic is actually delivering something users need. A streak counter can make a great product feel better. On its own, it just delays the churn.
Already have a product and not sure your engagement mechanics are doing real work? Impact Week is a focused, one-week audit of your existing experience. For $1,700, our team identifies what's genuinely driving return visits, what's quietly working against retention, and where a well-placed reward loop could add something real. You walk away knowing exactly what to fix and why.
Building something new and wanting behavioral design in the foundation? SolutionLab compresses the critical decisions into one week. A dedicated solutions architect and UX/UI designer prototype your core experience, pressure-test your key assumptions, and hand you a navigable product simulation before a single line of production code is written. At $6,900, it's the clearest way to avoid building fast in the wrong direction.
Ready to build and ship the full product? TractionLab is Designli's full-cycle development model. Designers and engineers work together inside a dedicated product team for the life of the project, so behavioral design isn't a checkbox at the end of a sprint, it's part of every decision from day one. Scope, timeline, and pricing are structured around what your product actually needs.
Not sure which fits where you are right now? That's a good reason to have a conversation before committing to anything.
FAQs
What are variable rewards in UX design?
A variable reward is a payoff for an action where the size, timing, or existence of the reward is unpredictable. Because anticipation itself triggers a dopamine response, unpredictable rewards produce more repeated behavior than a reward delivered the same way every time.
Is designing with variable rewards manipulative?
Not inherently. The difference comes down to intent: a reward loop that helps a user get better at something they came to do sits in a different category from one built purely to maximize time in the app regardless of whether the user benefits.
How do apps like Duolingo and TikTok use variable rewards?
Duolingo pairs a visible streak counter with unpredictable bonuses, so finishing a lesson always carries a bit of uncertainty even though the underlying habit is consistent. TikTok's recommendation algorithm takes it a step further: it varies each person's feed individually, in real time, based on what has kept that specific person watching before.
Variable Rewards for SaaS
Streaks, badges, and unpredictable payoffs work as a mechanism. That part isn't really in question; it's the whole point of everything above. Whether they belong in your SaaS product is a different question, and it depends far more on whether you actually understand what your users need in the first place.
A reward loop bolted onto a product that doesn't solve a real problem just delays the churn. It doesn't fix anything. If your users aren't sure why they're paying you, a streak counter won't change that. It'll just push back the moment they figure it out.
The reverse is also true. If the need is real and the solution is genuinely clear, a well-placed variable reward can make a good product feel even better. It gives someone one more reason to open the app on a day they might have otherwise skipped. But that only works because the foundation underneath it is already solid. The reward is amplifying something real, not covering for something that isn't there.
So before building a mechanic like this into a SaaS product, ask the more basic question first: do you know exactly what your users need help with, and is it obvious to them how you're helping? If the answer is yes, a variable reward is worth testing, carefully and with the same caution this article has recommended throughout. If the answer is no, that's the problem to solve first, not the reward loop.
Want to talk through how to build reward loops that serve your users? Get in touch.
Related Topics:
Behavioral Design is the Future of UX
