A punch card tells members exactly when the free coffee comes. A variable reward keeps them guessing. That guess has decades of research behind it, and it sits close to gambling. Loyalty teams that use it well keep the base reward fixed and layer chance on top, then measure both against a holdout.
What are variable rewards?
A variable reward comes after a number of actions, or a wait, that the person cannot predict. It comes from operant conditioning, the study of how results shape what we do. B.F. Skinner's lab sorted rewards into schedules. Each one is fixed or variable. Each counts either actions, called ratio, or time, called interval (OpenStax Psychology 2e).
Variable ratio is the schedule most people mean. The number of actions needed for a reward changes each time. OpenStax calls it the strongest of the partial schedules and uses gambling as its example. Of the four schedules, it gets the most responses. It is also the slowest to fade when rewards stop (OpenStax Psychology 2e).
| Schedule | How the reward arrives | Typical response pattern | Loyalty program example |
|---|---|---|---|
| Fixed ratio | After a set number of actions | High rate, short pause after each reward | Tenth coffee free on a punch card |
| Fixed interval | After a set amount of time | Long pause after each reward, rising toward the next | A monthly or birthday reward |
| Variable ratio | After an unpredictable number of actions | High and steady, little pause | An instant-win game piece with each purchase |
| Variable interval | After an unpredictable amount of time | Moderate and steady | A bonus day the program picks for each member |
The response patterns in the table come from OpenStax's summary of the four schedules. The loyalty column maps each one to a common program mechanic.
Why do unpredictable rewards motivate people?
Part of the answer is in the brain's reward system. In 1997, Wolfram Schultz, Peter Dayan and Read Montague described dopamine neurons whose output "apparently signals changes or errors in the predictions" of future rewards (Schultz, Dayan and Montague, 1997). A reward that comes as expected teaches little. A reward that beats the forecast stands out.
A 2003 study went further. Uncertainty peaks when a reward has a 50% chance. Schultz's team found a dopamine response that rose with uncertainty, "a gradual increase in activity until the potential time of reward" (Fiorillo, Tobler and Schultz, 2003). Not knowing has its own pull.
People show the same pull in their choices. Luxi Shen, Ayelet Fishbach and Christopher Hsee found that people invest more effort, time and money to earn an uncertain reward. In their example, a 50% chance at $2 and a 50% chance at $1 beat a sure $2. The effect held only when people focused on the chase, and it vanished when they focused on the prize itself (Shen, Fishbach and Hsee, 2015).
Do variable rewards work better than fixed rewards?
Sometimes, and the field trials show where. In a 2008 trial, 57 adults with obesity were split into three groups. Controls had monthly weigh-ins. The other two groups had a lottery or a deposit contract. Over 16 weeks, the lottery group lost a mean of 13.1 pounds, against 3.9 pounds for controls. Seven months in, the gap was no longer statistically significant (Volpp and others, JAMA, 2008).
A 2016 trial of 281 employees tested daily step goals. With a lottery reward, people hit the goal on 36% of days. Controls hit it on 30%, and the gap was not significant. Only the group that stood to lose money already given to them did better than controls. When the rewards stopped, steps fell in every incentive group (Patel and others, Annals of Internal Medicine, 2016).
So chance can help, and framing can matter more. Gains also fade once rewards end. For a loyalty program, that argues for testing a game against a plain bonus of the same expected cost before rolling it out.
When do variable rewards become gambling?
Paid chance is where regulators look. Loot boxes are paid items in video games with random contents. A self-selected survey of 7,422 gamers found a link between loot box spending and problem gambling. The link was stronger than for other paid in-game items (Zendle and Cairns, 2018). The UK government's 2022 response cited 15 studies that found "a stable and consistent association" between loot box use and problem gambling. It chose industry measures over new law, and said children should not buy loot boxes without parental approval (UK government, 2022).
Promotions that use chance usually keep a free way in. In the US, sweepstakes sent by mail must state that no purchase is necessary to enter (39 U.S.C. 3001). McDonald's prints the same line on its Monopoly game. Rules differ by state and country, so legal review belongs in the launch plan.
Game design can borrow the wrong trick, too. OpenStax notes that near misses, almost winning, also light up the brain's reward centers (OpenStax Psychology 2e). A loyalty game that stages fake near misses is copying a slot machine. Leave it out.
How do loyalty programs use variable rewards?
McDonald's Monopoly is a live example. Its 2026 Monopoly game starts October 6 and lives in the app. Players must register for the game and opt in to MyMcDonald's Rewards. A game piece can pay "Anywhere from 1,500 to 7,000 MyMcDonald's Reward Points." The stated odds are "1:5 odds to win, mostly food prizes," and "No purchase necessary" (McDonald's, 22 September 2026).
Variable timing works too. Starbucks picks the Double or Triple Star Days for its Green members. Its terms say each one is "personalized and selected by Starbucks for you" (Starbucks Rewards terms). Gold and Reserve members get extra Double Star Days. The member never knows which week the bonus lands.
Three design rules follow. Keep the base earn rate fixed, so members trust the program. Put chance on the extras, like a game piece or a surprise bonus day. And price the game by its expected cost, the way finance prices any other bonus. Our entries on instant-win mechanics and gamification cover the formats.
For illustration, a program pays a fixed 200 points per visit, worth $2.00 at one cent a point. It tests a game instead: one visit in five wins, and wins average 800 points. The expected cost is 160 points, or $1.60 a visit. Run both against a random holdout for eight weeks. If visits hold, the game saves 40 cents a visit. If visits fall, the fixed bonus was doing work the game cannot.
How GRAVTY handles variable rewards
GRAVTY's airline page lists "Challenges, gamification, sweepstakes and raffles" as ways to keep members coming back (GRAVTY for airlines). Loyalty managers write bonus rules in the patented Visual Rules engine. They test the rules in a simulation first. Then they put them live with no IT ticket. The game and the base earn rate sit in the same rules.
At Deutsche Telekom, Magenta Moments keeps members engaged "between invoices" with partner offers, games and leaderboards. These ship in the base platform. The program runs with 4,500 partners across 9 countries (GRAVTY for telecom). AI-Trust flags odd earn and redemption behavior as it happens. So a prize that pays out too often shows up while the game runs.
Frequently asked questions
What is an example of a variable reward?
A slot machine is the classic example of a variable reward. OpenStax uses gambling to show the variable ratio schedule, since the player never knows which pull will pay (OpenStax Psychology 2e). In loyalty, McDonald's 2026 Monopoly game is one. A game piece can pay from 1,500 to 7,000 MyMcDonald's Rewards points (McDonald's, 2026).
Are variable rewards addictive?
Variable rewards make habits hard to break, and the link to harm shows up in paid forms that work like gambling. OpenStax notes that recent rat studies did not back Skinner's idea that these schedules alone cause pathological gambling (OpenStax Psychology 2e). Loot box spending, though, was linked to problem gambling in a survey of 7,422 gamers (Zendle and Cairns, 2018).
What is the difference between variable ratio and variable interval?
A variable ratio reward depends on how many tries it takes, and a variable interval reward depends on time. Ratio rewards pay after an unknown count of tries, as with a slot machine. Interval rewards pay after an unknown wait, as with a surprise spot check that pays the whole crew a bonus. OpenStax notes that ratio schedules get high, steady effort. Interval schedules get a moderate, steady pace (OpenStax Psychology 2e).
What is intermittent reinforcement?
Intermittent reinforcement means a behavior is rewarded only some of the times it happens. Its other name is partial reinforcement. Trainers often reward every time at first, since that teaches a new habit fastest. Once the habit is set, they reward only some of the time (OpenStax Psychology 2e). Variable rewards are one kind of it.
Do variable rewards keep working after the program stops them?
The evidence says the gains fade once rewards stop. In a 2016 step-count trial, steps fell in every incentive group after incentives ended (Patel and others, 2016). In a 2008 weight-loss trial, the lottery group's lead was no longer statistically significant at seven months (Volpp and others, 2008). Plan variable rewards as a standing part of the program.
