A $10 coupon feels different from $10 in a wallet, and a loyalty point feels different from a refund of the same value. That is mental accounting. People sort money into separate accounts and give each account its own rules. The idea earned Richard Thaler a Nobel prize, and it shapes how members spend, save and redeem.

What is mental accounting?

Mental accounting is the set of mental steps people use to organize, track and judge money. Richard Thaler introduced the model in 1985, in Marketing Science. It joins the prospect theory value function to a new idea, transaction utility, and adds household budgeting (Thaler, 1985). In 1999 he reviewed what researchers had learned since. Some explainers date the term to that review, but the model is older.

The Nobel committee put it plainly when it honored Thaler in 2017. People simplify money choices by creating separate accounts in their minds. They focus on the narrow effect of each choice and not its overall effect (Nobel Prize press release, 2017).

What are the three parts of mental accounting?

Thaler’s 1999 review names three components. Each one breaks the economic rule that money is fungible, meaning a dollar is a dollar wherever it sits (Thaler, 1999). The table shows each part and where it appears in a loyalty program.

ComponentWhat it meansWhere it shows in a program
How outcomes are codedGains and losses are perceived and evaluated one decision at a timeA bonus feels like a gain on top of a purchase; an expiry feels like a separate loss
Which accountMoney is labeled by source and use, and spending is held to budgetsStarbucks says Stars are promotional, have no cash value and cannot be redeemed for cash (terms)
How often accounts are balancedAccounts can be balanced often or rarely, and drawn narrowly or broadlyDelta resets Medallion Qualification Dollars each calendar year (Delta)

What are classic examples of mental accounting?

The earliest evidence is a pair of questions from Amos Tversky and Daniel Kahneman. In the first, people had decided to see a play and found at the door that they had lost a $10 bill. 88% of 183 respondents would still buy a ticket. In the second, people found they had lost the ticket they had already bought. Only 46% of 200 respondents would buy another (Tversky and Kahneman, 1981). The authors argued that a second ticket gets booked to the same account as the first, so the show seems to cost $20.

A second pair used a jacket and a calculator. People were asked whether they would drive 20 minutes to save $5 on a calculator. 68% would when the calculator cost $15. Only 29% would when it cost $125 (same paper). The $5 is the same, but it sits in a different account.

Is mental accounting real?

Mostly yes, with some classics weaker than their reputation. In 2025 Mengfei Li and Gilad Feldman replicated 17 problems from Thaler’s 1999 review with 1,007 participants. They found support for 11, mixed support for 3 and no support for 3. The lost-ticket result held in direction, with 90% buying after losing a bill and 77% after losing a ticket. The calculator result did not, since 11% and 7% would make the trip (Li and Feldman, 2025).

Field data point the same way. Justine Hastings and Jesse Shapiro studied 10.5 million gasoline purchases by 61,494 households. A $1 rise in the price of gasoline raised a typical household’s chance of buying regular gas by 1.4 percentage points. An equal loss of income would have moved it by less than a tenth of a point. Households treated money for gas as separate from other money (Hastings and Shapiro, 2012).

Milkman and Beshears studied customers of an online grocer who used $10-off coupons. They compared each customer’s coupon orders with her other orders. Grocery spending rose $1.59 when a coupon was used. The extra went to items the customer did not usually buy (Milkman and Beshears). Spending here means the groceries’ price before the coupon.

How does mental accounting show up in loyalty programs?

A program is a set of accounts. Starbucks keeps dollars on a Starbucks Card and Stars in Rewards, and it links them with a bonus. Members who digitally reload $30 receive ten Bonus Stars, and a $50 reload earns 25 (Starbucks Rewards Terms). The reload bonus pulls cash into the Stars account on purpose.

Time brackets work the same way. Delta counts Medallion Qualification Dollars on a calendar-year clock and resets them each calendar year (Delta Air Lines). Members judge their progress against a yearly account, and the reset gives them a fresh one.

Three operating rules follow. First, a voucher acts as a small windfall, and the grocer data show a lift on top of the discount. Second, the clock matters. A yearly status year invites members to check their balance once a year, while a monthly offer invites a monthly check. Third, labels matter. The same value described as points, cash back or a credit will land in a different account.

How should a program present gains and losses?

Thaler’s 1985 paper proposed four editing rules, and the 2025 replication supported all four with weaker effects than the originals. People prefer to segregate gains and to integrate losses. They also prefer to integrate a smaller loss with a larger gain, and to segregate a small gain from a large loss (Li and Feldman, 2025). In a program, those rules become four presentation choices. Show a bonus as its own line and not folded into the base earn. Combine small charges into one amount. Put an expiry notice beside a larger gain, such as this month’s earn. Give a small perk its own moment when the member faces a large cost. The replication authors call the wider hedonic editing effect only partially supported, so treat these rules as starting points for tests. They are applications and not tested loyalty results.

How can a loyalty team measure it?

Compare each member with herself. For illustration, a program sends a $10 reward to 5,000 members and 1,500 redeem it. A team that matches each redeemer’s reward orders to her own other orders sees whether baskets grow. If the lift matches the grocer’s $1.59, it adds $2,385 in basket value across the 1,500 orders, against $15,000 in rewards.

That gap is the lesson. A mental accounting lift is real and small. The reward must earn its cost from the orders it brings back and the members who stay longer. Partner funding can cover part of it. Test it against a holdout that gets no reward, and read the result on margin.

How GRAVTY handles mental accounting

Whether value arrives as a voucher, points or status is a rule choice. The GRAVTY technology page says business teams configure rewards, recognition and redemption for any segment without code. Agentic AI Compass lets a team compare results across any segment, which is how a test of two reward forms gets read. Our guide to loyalty program ROI covers the holdout method.

Frequently asked questions

What is an example of mental accounting?

The lost ticket and lost bill pair is the classic example. 88% of respondents would still buy a $10 play ticket after losing a $10 bill, but only 46% would buy a second ticket after losing the first (Tversky and Kahneman, 1981). The loss is the same $10, and people file it in different accounts.

Who came up with mental accounting?

Richard Thaler came up with it, and he laid out the model in 1985 (1985 paper). He won the 2017 Nobel prize in part for it (Nobel press release). Tversky and Kahneman had shown the effect in 1981 (1981 paper).

Is mental accounting a bias?

Yes, in the sense that it breaks the rule that money is fungible. Thaler says each component of mental accounting violates fungibility (Thaler, 1999). People also use accounts as a budgeting aid, so the effect can help as well as hurt.

How does mental accounting affect loyalty programs?

It makes points, vouchers and cash feel like different money. Starbucks says Stars have no cash value and cannot be redeemed for cash (Starbucks Rewards Terms). In grocery data, redeeming a $10-off coupon raised spending by $1.59 (Milkman and Beshears). Members spend a voucher differently from cash.

Does mental accounting replicate?

Mostly. A 2025 replication of 17 problems from Thaler’s 1999 review found support for 11, mixed support for 3 and no support for 3 (Li and Feldman, 2025). The lost-ticket pattern held and the calculator result did not.