The Variable Reward
Make rewards uncertain to build habits that resist extinction
- Difficulty
- Easy
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 90%
B.F. Skinner's rats learned to press a lever for a fixed sugar drop, but the habit decayed fast once the sugar stopped. A second group whose presses paid out variably, sometimes nothing, sometimes one, two or three drops, averaging one, kept pressing long after the reward was cut. Skinner concluded that variable, uncertain rewards embed habits far more deeply than fixed ones, and later work by Ayelet Fishbach showed the same holds for humans. The practical translation is to swap fixed loyalty rewards (buy ten coffees, get one free) for a variable schedule (staff give away roughly ten percent of coffees at random). The average cost is identical, but the unpredictability makes the habit stronger. The same mechanism drives social media notifications and McDonald's Monopoly.
Origin
Rooted in B.F. Skinner's 1930s operant-conditioning box experiments and extended to humans by Ayelet Fishbach, as recounted by Richard Shotton.
Core principles
- 01Uncertain rewards embed habits far more deeply than fixed ones
- 02Variability, not size, is what keeps behavior alive after rewards stop
- 03The same total spend delivers more pull when made unpredictable
- 04The mechanism powers social media, gambling and loyalty schemes alike
How to run it
- 1
Pin down the target behavior
Decide the exact repeat action you want to reinforce, such as a returning purchase or a daily app open.
- 2
Audit your current reward for certainty
Most loyalty schemes offer a fixed, transactional payoff. Note where the reward is fully predictable, because that is the weakness.
- 3
Introduce variability at constant average cost
Redesign so the reward arrives on an unpredictable schedule while the average payout stays the same, so spend does not rise.
Pro tip Give staff the option to hand out roughly 10% of items free rather than a rigid ten-to-one punch card.
- 4
Let people feel the uncertainty
Make the not-knowing salient, the moment of opening the app or peeling the sticker, so anticipation itself drives the return.
- 5
Guard against gaming and context effects
Test in your specific setting; unpredictable rewards can trigger unintended behavior, and effects vary by context.
Watch out Stockholm's speed-camera lottery cut average speeds but some drivers circled the block to re-enter the draw.
In the wild
A standard scheme promises a free coffee after ten purchases, a fixed and certain reward. Applying Fishbach's principle, the shop instead gives staff discretion to comp roughly ten percent of coffees at random. One customer might go twenty times without a freebie; another wins every fifth visit. Average cost is unchanged but the uncertainty makes the habit far stickier.
→ Stronger repeat behavior for identical reward spend.
Opening your phone, you never know how many notifications, DMs or retweets await. Shotton argues that if a tweet's reception were fully known in advance, the pull of the platform would have faded; it is the uncertainty over whether you get two retweets or fifty that keeps people returning.
→ Uncertain social feedback sustains compulsive daily use.
Common mistakes
Offering only fixed, certain rewards
A predictable ten-for-one reward builds a shallow habit that dies when the reward stops; adding variability at the same cost deepens it.
Ignoring how context invites gaming
Unpredictable rewards can be exploited, as when Stockholm drivers looped the block to re-enter a speed-camera lottery, so pilot before scaling.
Is it for you?
Best for
Loyalty programs, apps and any business trying to build durable repeat behavior.
Not ideal for
Contexts where customers demand guaranteed, transactional value and dislike surprise.
From the transcript
“if you want to create a habit, far better to reward the behavior that you want with an uncertain and variable reward rather than a…”
“you can make that scheme far more powerful without spending any more money by adding an element of variability”
From the episode
8 Fascinating Psychological Biases - Richard Shotton - #592
Richard Shotton