A stamp card is a lab schedule with a logo on it. Buy nine drinks and the tenth is free, so the card pays on a fixed count. The science behind that rule is close to a century old. It also explains why members slow down right after a reward, and what happens to buying when a bonus ends.

What is operant conditioning?

Operant conditioning is learning from what happens after you act. A 2003 review in the Annual Review of Psychology puts it in one line: operant behavior is behavior “controlled” by its consequences. B. F. Skinner coined the term in 1937 (Staddon and Cerutti). He chose the word because the act works on the world to get a result (Skinner, 1953).

The idea starts with Edward Thorndike’s law of effect. In 1911 he wrote that a response “closely followed by satisfaction” becomes more firmly tied to the situation, so it is more likely to recur. A response followed by discomfort grows weaker (Thorndike, 1911). Skinner then built a box to measure it. A rat pressed a lever or a pigeon pecked a disk, food came out, and a recorder counted each response (OpenStax Psychology 2e).

Operant learning differs from the kind Pavlov studied, where a reward is paired with a signal. In operant learning the reward depends on what the animal does (Skinner, 1953). A loyalty program is the operant kind. Nothing is paid until the member acts.

What are the four types of operant conditioning?

The four types come from two choices. Do you add something or take it away? Do you want the behavior to rise or fall? Here positive means adding and negative means taking away. Reinforcement raises a behavior and punishment lowers it (OpenStax).

  • Positive reinforcement adds a thing the member wants, such as points for a purchase. The act it follows grows more likely.
  • Negative reinforcement takes away a thing the member wants to avoid. The act still grows more likely, since it ends the bad thing.
  • Positive punishment adds a thing the member does not want, such as a fee. The act it follows grows less likely.
  • Negative punishment takes away a thing the member values, such as a perk. That too makes the act less likely.

Points expiry rules often run on negative reinforcement. Starbucks Green members lose Stars six months after the month they were earned. A purchase, a redemption or a digital reload of $30 or more in the month before keeps them alive for another month (Starbucks Rewards Terms). The purchase that saves the Stars is rewarded by removing a loss.

What are the schedules of reinforcement?

A schedule is the rule for when the reward comes. A reward for every act is called continuous reinforcement, and it is the fastest way to teach a new act. Partial schedules pay for only some acts. Each is fixed or variable, and each counts either acts or time (OpenStax).

ScheduleWhen the reward comesWhat behavior looks likeA loyalty version
Fixed ratioAfter a set number of actionsA high rate, with a short pause after each rewardCaffè Nero pays the 10th drink after nine stamps (Caffè Nero)
Variable ratioAfter a number of actions that keeps changingA high, steady rate with little pause; the slowest to fadeThe 2026 Monopoly game at McDonald’s, with 1 in 5 odds of an instant win (Official Rules)
Fixed intervalAfter a set amount of timeA long pause after the reward, then a riseA once-a-year birthday reward, such as the one in Starbucks Rewards (Starbucks)
Variable intervalAfter an amount of time that keeps changingA steady rate at a middle paceBonus point days sent on no set date

The patterns in the table come from OpenStax. It rates variable ratio the most productive schedule and the hardest to stop. Fixed interval is the least productive and the easiest to stop (OpenStax). Skinner tied the variable-ratio schedule to slot machines and roulette wheels (Skinner, 1953).

A program that adds prizes of chance borrows that schedule, so the guardrails matter. To play the McDonald’s game, a customer opts in to MyMcDonald’s Rewards. Each person can enter up to eight game codes a day. A game piece can also be requested by mail with no purchase (Official Rules).

Why do members slow down right after a reward?

Fixed-ratio schedules build in a pause. Skinner saw a very low chance of a response just after each reward. In piecework, he wrote, low morale or low interest shows up most often just after a unit of work is done (Skinner, 1953).

The same dip shows up in a coffee card. Ran Kivetz, Oleg Urminsky and Yuhuang Zheng tracked 110 members who finished one card and started a second. The last two gaps between purchases on the first card averaged 2.2 and 2.1 days. The first two gaps on the new card averaged 3.1 and 2.7 days (Kivetz, Urminsky and Zheng, 2006). The authors call it postreward resetting.

For illustration, a café program sees 4,000 finished cards a month. Say each one is followed by that same slowdown. The first two visits on the new card take 5.8 days in place of 4.3, a loss of 1.5 days. At one visit every 2.15 days, that is about 0.7 visits per card, or roughly 2,790 visits a month. At $3.50 of margin a visit, the pause costs about $9,800 a month. A bonus on the first visit of the new card is the obvious test, against a random holdout.

What happens when the rewards stop?

The act fades, and psychologists call that extinction. Skinner gave a plain case: when our calls go unanswered, we stop calling (Skinner, 1953). Partial schedules slow the fade. He told of a pigeon, trained on a special schedule, that pecked more than 10,000 times after its food stopped. A reward for every peck, he wrote, never leads to anything like it.

Michael Rothschild and William Gaidis warned marketers in the Journal of Marketing in 1981. If buying rests mainly on the incentive, they wrote, taking the incentive away leads to extinction (Rothschild and Gaidis, 1981). Their advice was to tie the buying to the product. The incentive’s job is to shape that buying, step by step.

Their shaping example starts with a free sample that carries a large coupon. Each later purchase holds a smaller one. The incentive fades until the product alone keeps people buying (Rothschild and Gaidis).

A devaluation is a change to the schedule. Skinner flagged a risk of ratio pay: as output rises, the payer moves to a larger ratio. He also found a limiting ratio past which the behavior stops (Skinner, 1953). Raising the points needed for a reward moves members toward that limit.

Rewards also carry a cost in drive. Edward Deci, Richard Koestner and Richard Ryan pooled 128 studies in 1999. Expected, tangible rewards cut how much people kept at an interesting task in a later free period. Praise and other positive feedback raised it (Deci, Koestner and Ryan, 1999). Paying points for something members like to do, such as writing a review, needs a test first.

How do loyalty programs use operant conditioning?

Points work as tokens. A secondary reinforcer has no value of its own and gets it from what it buys, and token economies are built on them (OpenStax). Tiers change the rate. Starbucks Green members earn 1 Star per dollar, Gold members 1.2 and Reserve members 1.7 (Starbucks Rewards Terms).

Timing matters as much as size. The textbook rule is that the consequence comes soon after the response (OpenStax). Rothschild and Gaidis put mail-in premiums at four to six weeks from purchase to reward. They argued such premiums end up rewarding the trip to the mailbox more than the buying (Rothschild and Gaidis). Points that post at the till close that gap.

Every reinforcement claim needs a comparison group. A reward that lands on members who would have bought anyway trains nothing new. Control group testing sets up the holdout, and our punch card entry covers stamp card economics.

How GRAVTY handles operant conditioning

A reinforcement schedule is program logic, and the GRAVTY technology page says “Patented Visual Rules let business teams change program logic”. Visual Rules is a patented no-code rules engine. The team that owns the stamp count or the bonus day can change it.

For retail programs, lji.io describes AgenticTest in its own words. It runs a “Live A/B simulation pre-launch” and “watches lift in flight” (GRAVTY for retail). Prizes of chance draw abuse. GRAVTY AI-Trust uses machine-learning models to flag anomalous earn and redemption behavior as it happens.

Frequently asked questions

What is the difference between operant and classical conditioning?

Classical conditioning pairs a reward with a signal, while operant conditioning makes the reward depend on what the learner does. Skinner drew the line in 1953. In Pavlov’s work the reward is paired with a signal. In operant work it hangs on what the animal does (Skinner, 1953). Points for a purchase are operant.

Who discovered operant conditioning?

B. F. Skinner coined the term in 1937, and he built on Edward Thorndike’s law of effect (Staddon and Cerutti). Thorndike set out that law in 1911. Acts that bring satisfaction grow more likely to recur (Thorndike, 1911). Skinner’s box made the effect easy to count.

Is negative reinforcement the same as punishment?

No, because negative reinforcement makes an act more likely and punishment makes it less likely. Negative reinforcement takes away a thing the learner wants to avoid. Punishment adds a thing they do not want, or takes away a thing they value (OpenStax). A purchase that stops points from lapsing is a case of the first.

Which reinforcement schedule is hardest to break?

The variable-ratio schedule is the hardest to break. The OpenStax text calls it the most productive schedule and the slowest to fade, and it uses gambling as the example (OpenStax). Fixed interval is the easiest to break.

What is an example of operant conditioning in marketing?

Coupons that shrink with each purchase are a classic example. Rothschild and Gaidis described a free sample with a large coupon, then smaller coupons inside each later purchase, until the product alone keeps people buying (Rothschild and Gaidis, 1981). Psychologists call this step-by-step method shaping.