Gift Card Breakage - The money left behind

Dividing this year’s breakage by this year’s loads can be useful as a clearly labeled accounting-intensity ratio. It is generally not a cohort breakage rate: the recognized amount can relate to older cards while the denominator contains newly sold cards whose redemption is barely underway.

The reverse mistake appears during a sales slowdown. Recognition from older cohorts can remain high while new loads shrink, mechanically raising the ratio. A change in that ratio does not by itself show that customers became more forgetful. Loyalty balances, refunds and promotional credits can further contaminate the denominator.

A sound cohort estimate matches cards issued in the same period and follows their redemptions to a common age, while separating refunds, cash returns, transfers and legally remittable balances. The same calendar month can contain mature Christmas cards and fresh promotional rewards, each with a very different remaining life.

The data behind the story

F17. Unchanged recognition can produce very different apparent rates

Measure: %.

%

Explore the data table

← Scroll left or right to see all columns →

F17. Unchanged recognition can produce very different apparent rates (%)

Category	Recognized breakage / current loads
Year A	5.0%
Year B	2.5%
Year C	10.0%

T15. Illustrative denominator trap

[M1]

How to estimate breakage without declaring living cards dead

The data should be organized by issuance cohort and age since issuance. Start with money, not just card counts: a mostly exhausted $100 card and an untouched $100 card are not equivalent exposures. Preserve partial redemptions and distinguish first use from final depletion.

Milliman describes multiplicative development, additive future-redemption estimates and mature-cohort analogues. It explicitly rejects a universal curve: its approximately 10% tail illustration is hypothetical. Different channels, denominations and disruptions can change both timing and ultimate use. [R1]

For a more advanced model, a mixture-cure formulation can separate eventual non-use from the timing of use among eventual redeemers. That sophistication does not eliminate the central identification problem: a card that has not been used by the end of a short sample may be a slow redeemer, not a permanent non-redeemer. Extrapolating a long tail requires assumptions, mature history and sensitivity analysis.

For partial spending, a multi-state or marked-event model is more informative than a first-redemption survival curve alone. Track untouched, partly used, exhausted, refunded, remitted and otherwise resolved states. Preserve migration between cards or wallets so a technical replacement is not mistaken for a new customer obligation. These are proposed analytical practices, not claims that a particular company uses them.

The data behind the story

F18. Slow redemption and higher ultimate breakage are different stories

Measure: % of original face redeemed.

%

Explore the data table

← Scroll left or right to see all columns →

F18. Slow redemption and higher ultimate breakage are different stories (%)

Category	Fast: 8% ultimate breakage	Slow: 8% ultimate breakage	Fast: 20% ultimate breakage
1 months	14.12%	4.97%	12.28%
3 months	36.20%	14.12%	31.48%
6 months	58.16%	26.08%	50.57%
12 months	79.55%	44.77%	69.17%
24 months	90.31%	67.75%	78.53%
36 months	91.77%	79.55%	79.80%
60 months	92.00%	88.72%	80.00%
84 months	92.00%	91.13%	80.00%

T16. Models answer different questions

[R1] [M1]

The measurement system an advanced investigation needs

The decisive dataset is a reconciled ledger linking issuance, funding, redemption, adjustments and final disposition. It should permit both a financial roll-forward and a customer-outcome analysis. A daily balance export without event history cannot establish whether residuals vanished through legitimate redemption, migration, refund, fees or data loss.

Use separate tables for economic face value and accounting carrying value. Give every record a currency and retain original currency until a transparent conversion rule is chosen. A bank settlement, a merchant order and a gift-card redemption may be three records of one economic event, not three separate uses of money.

Validation should begin with conservation: opening face plus loads and reinstatements minus redemptions, refunds, valid fees and transfers out must equal closing face after explained adjustments. Then reconcile the accounting roll-forward separately, including breakage and estimate revisions. Failures in either identity should block publication of an aggregate rate.

T17. Minimum analytical fields

T18. Quality tests that change the conclusion

A small estimate change can create a large income effect

For a simple eligible cohort, let F be issued value, b the expected ultimate breakage fraction and R cumulative redemptions. Expected ultimate redemption is F(1-b); cumulative proportional breakage is R × b/(1-b), subject to the appropriate recognition constraints and an ultimate cap. Current-period recognition is the change from the previously recognized cumulative amount.

In the illustrative $1m cohort below, $460,000 has been redeemed. Raising the estimated breakage rate from 8% to 12% increases cumulative recognized breakage from $40,000 to about $62,727, without another customer transaction. Reducing the estimate to 4% lowers it to about $19,167. That is why changes in estimation require an explanation separate from sales performance. [M1]

The derivative R/(1-b)^2 describes local sensitivity to b in this simplified formula. It is a mathematical sensitivity, not a probability or permission to book an amount. Actual recognition also depends on the applicable standard, legal entitlement, data quality and constraints on reversal risk.

The data behind the story

F19. Recognition moves even when redemptions do not

USD

Explore the data table

← Scroll left or right to see all columns →

F19. Recognition moves even when redemptions do not (USD)

Category	Cumulative recognized breakage
2%	9,387.76 USD
4%	19,166.67 USD
6%	29,361.70 USD
8%	40,000.00 USD
10%	51,111.11 USD
12%	62,727.27 USD

T19. Illustrative estimate sensitivity at the same redemption level

F=$1m; R=$460,000. Rounded display only; underlying calculations retain precision. [M1]

Float is a timing benefit, not another name for breakage

Breakage concerns principal that will not be redeemed. Float concerns the interval before cash must be spent or settled. A fully redeemed program can still have a financing benefit; a program with high breakage may have little usable float if funds are restricted or the economics belong to another party.

For an analytical estimate, separate average funds actually available for investment, the realized net yield, custody restrictions, settlement timing and who receives interest. Multiplying every outstanding consumer balance by a headline interest rate assumes answers to all of those questions.

The scenario below illustrates that distinction. It uses a hypothetical $100m average balance, investable shares of 0%, 50% and 100%, and assumed net annual yields. It excludes taxes and operational costs. Nothing here estimates Starbucks, a bank issuer or any named program’s interest income. [M1]

The data behind the story

F20. Available funds determine the financing benefit

USD million per year

Explore the data table

← Scroll left or right to see all columns →

F20. Available funds determine the financing benefit (USD million per year)

Category	50% available	100% available
0%	0.0 USD million per year	0.0 USD million per year
1%	0.5 USD million per year	1.0 USD million per year
2%	1.0 USD million per year	2.0 USD million per year
3%	1.5 USD million per year	3.0 USD million per year
4%	2.0 USD million per year	4.0 USD million per year
5%	2.5 USD million per year	5.0 USD million per year

T20. Illustrative annual float benefit on a $100m average balance

[M1]

For reward buyers, non-use can mean paying for an outcome that never happened

For a merchant selling its own paid gift card, breakage can contribute to financial results. For an employer or promotional sponsor buying a reward, the same non-use may mean its intended benefit never reached the recipient. The cash recipient and the party seeking engagement are not necessarily the same business.

A reward also has more than one drop-off point: approved, delivered, claimed, activated, first spent and fully spent. A dashboard that stops at email delivery cannot measure realized reward value. An unclaimed reward link is not automatically a funded card balance, and a funded unused card is not automatically revenue for the distributor.

Ask when the sponsor is charged, whether unused value is returned, who holds the funds, what happens after expiry, and how reports distinguish non-claim from post-activation residuals. Those are contract questions. Marketing labels such as no breakage or pay on use require an operational definition before they can be compared.

The data behind the story

F21. Unreturned unused value raises the sponsor’s cost per redeemed dollar

USD per redeemed USD

Explore the data table

← Scroll left or right to see all columns →

F21. Unreturned unused value raises the sponsor’s cost per redeemed dollar (USD per redeemed USD)

Category	Sponsor cost / redeemed value
0%	1.020 USD per redeemed USD
5%	1.074 USD per redeemed USD
10%	1.133 USD per redeemed USD
20%	1.275 USD per redeemed USD
30%	1.457 USD per redeemed USD

T21. Questions that reveal the reward economics

Recovering value and separating fraud

A merchant should measure the cost of recovery as well as its benefit

Reducing breakage can improve customer outcomes while having an ambiguous short-term profit effect. Redemption consumes goods or services; it can also drive incremental spending and later contribution. The right comparison is incremental contribution against a credible no-intervention counterfactual, not total sales made by everyone who used a card.

Consumer preferences can help choose which recovery measures to investigate. In another Accenture 2024 US table published by Statista, 36% selected simpler balance checks, 35% regular reminders for unused balances, and 32% easier redemption across online and in-store channels. More versatile use also drew 32%. Statista reports 1,501 respondents for this question, a different base from the 483 unused-balance respondents in F07. [ST04]

The data behind the story

F22. Consumers ask for easier access to existing value

Requested improvements to the gift-card experience, rather than measured effects of a recovery campaign.

% of surveyed US consumers

Explore the data table

← Scroll left or right to see all columns →

F22. Consumers ask for easier access to existing value (% of surveyed US consumers)

Category	Selected this improvement
Simpler balance checks	36 % of surveyed US consumers
Reminders for unused balances	35 % of surveyed US consumers
Easier online/in-store redemption	32 % of surveyed US consumers
More versatile use	32 % of surveyed US consumers
More personalization	30 % of surveyed US consumers
Option to upgrade the gift card	26 % of surveyed US consumers
Clearer usage instructions	23 % of surveyed US consumers
Better technical support	23 % of surveyed US consumers

The findings provide candidates for action, not a revenue forecast. A reminder may bring a planned purchase forward, recover value that would otherwise remain idle, or prompt a purchase that would have happened anyway. Better balance visibility may resolve an access problem without producing an additional sale. Only a comparison that measures the intervention’s incremental effect can distinguish these outcomes; that is why the recovery calculation and design in T22–T23 remain necessary.

The companion article What people want when they check a gift card balance examines the balance-checking task and provides official checking routes for 20 brands and programs.

For a simplified closed-loop program, recover D of previously unredeemed face value. Let c be the variable cost fraction of fulfilling that value, U incremental cash spending caused by the intervention, m its contribution margin, H incremental future contribution and K campaign cost. Incremental contribution is mU + H - cD - K. This holds the original card-sale cash fixed and ignores tax, discounts and capacity constraints.

An illustrative recovery of $10,000, variable cost of 60%, margin of 40% on extra spending and $1,000 campaign cost requires $17,500 of incremental extra spending if there is no later contribution. With $5,000 of demonstrable future contribution, the requirement falls to $5,000. The calculation is an analytical break-even, not an industry benchmark. [M1]

T22. Illustrative recovery break-even

[M1]

T23. An experiment that can distinguish access improvement from unprofitable subsidy

A drained card and an abandoned card are opposite ledger events

Fraud can leave a customer with no usable value because someone else redeemed it. Breakage arises from value not redeemed. A customer complaint that a card has no balance cannot establish which occurred. The distinction matters both for restitution and for claims about issuer profitability.

The FTC describes Consumer Sentinel as an administrative collection of unverified reports, not a consumer survey. Its statistics therefore need their own reporting and coverage caveats. They should not be added to survey unused balances or recognized corporate breakage to create one grand total of gift-card waste. [R3]

Statista’s 2024 fraud tables illustrate the distinction in monetary terms: they reproduce 41,120 reports and $212 million in reported losses from the FTC’s payment-method data. The original category is “Gift Card or Reload Card.” It is broader than retailer gift cards and does not isolate theft from a card that its owner intended to keep. [ST05] [ST06] [ST08]

T24. Reported fraud involving a payment method is not breakage

The FTC limits the dollar amounts in this table to reports of losses from $1 through $999,999; the report count is a separate payment-method measure. Consumer Sentinel is not a probability sample or a complete census, and the table should not be treated as a US-only national estimate. A card used to pay a scammer may have been fully redeemed. Such a loss belongs in a fraud analysis, not in the same numerator as value that nobody redeemed. [ST08] [R3]

A proper investigation follows the card event log: activation, authorized and disputed transactions, fees, refunds and replacement links. It also asks who bore the loss. A reimbursed victim, a merchant chargeback and a program reserve expense describe different economic outcomes.

T25. Classifying what happened to the value

Accounting rules, consumer rights and enforcement

Which law applies depends on what was issued

A paid retailer card, an open-loop gift card, a promotional reward and a reloadable prepaid account can follow different rules. Open-loop acceptance alone is not an exemption from U.S. federal gift-card protections. Qualifying promotional/reward instruments have specific disclosure requirements; their label is not sufficient. [L06] [L07]

For covered U.S. instruments, the federal funds-expiry floor is generally five years from issuance or last load, as applicable, or the later card expiry. Qualifying inactivity/service fees require a year of inactivity, disclosure and no more than one such fee a month. State protections may be stronger. The assertion that all dormancy fees were eliminated is inaccurate. [L06]

The following matrices are a selected-jurisdiction comparison, not a worldwide or fifty-state compliance inventory. Finite minimum-validity periods do not measure actual redemption speed and do not rank total consumer protection.

The accounting choice starts with the issuer's promise. An obligation to supply its own goods differs from an obligation to pay another merchant when a cardholder spends. The IFRS Interpretations Committee's 2016 decision turned on that payment obligation, even though the customer could not redeem the card for cash. [L04]

Small-balance rules also change the possible outcome of an unused card. California's test is a balance below $15, while Ireland's EUR1 remedy depends on a contract preventing later use and permits a replacement voucher. The amount alone does not identify the holder's remedy. [L08] [L15]

The data behind the story

F23. Selected minimum-validity rules range from three to nine years

years

Explore the data table

← Scroll left or right to see all columns →

F23. Selected minimum-validity rules range from three to nine years (years)

Category	Covered minimum
Australia	3 years
New Zealand	3 years
U.S. federal	5 years
Ireland	5 years
Massachusetts	7 years
New York	9 years

T26. Accounting scope is not universal

[L01] [L02] [L03] [L04] [L05]

T27. Selected U.S. cash-redemption and validity rules

[L08] [L09] [L12] [L13]

T28. Selected international rules

[L14] [L15] [L16] [L17] [L18] [L19] [L20] [L21] [L22]

T29. Who can claim unredeemed value?

The Supreme Court source explains general priority rules; its MoneyGram-specific holding is not a gift-card holding. [L10] [L11] [L01] [L02]

T30. Tax can run on a separate timetable

ILLUSTRATIVE federal income-tax timing, not tax due. Eligibility and election requirements matter; sales tax/VAT not modeled. [L24] [M1]

Enforcement shows why an expired balance may not be gone

In July 2026 Ireland’s CCPC announced that JD Sports would reissue 5,604 online cards, representing EUR246,859, after an incorrect one-year expiry. The announced completion deadline is 4 January 2027, after this study’s cutoff. This is an undertaking to restore value, not verification that restoration is complete or that the amount had been booked as breakage. [L23]

In New York, the Attorney General’s 2024 account describes more than $36m recovered from H&M in 2022 and a later settlement with Card Compliant concerning alleged avoidance of unclaimed-property obligations. The exact latter settlement is $4,375,000; the signed agreement says it is a compromise of disputed claims without admission of liability or wrongdoing. These are not criminal convictions or measured consumer breakage rates. [A06] [A07]

The common issue is control over the obligation. A software status change or an intercompany contract does not by itself establish the right to retain the money. A journalist must examine the actual operating arrangement, the applicable legal test and the procedural status of any enforcement claim.

T31. Two cases, two distinct outcomes

[L23] [A07]

National estimates and transparent disclosure

The national total remains a reporting problem

There is no defensible shortcut from a few prominent company disclosures to a national breakage total. The sample is selected by disclosure availability, contains different product scopes and fiscal periods, and excludes many private companies. A sum can describe the selected disclosures; it cannot establish market coverage.

An earlier attempt illustrates the difficulty. A 2020 SSIR viewpoint described an estimated $4.22bn for selected companies in 2017, split into $381m from full disclosers, $755m from hybrid disclosers and $3.085bn from nondisclosers. Most of that estimate was therefore not a directly disclosed standalone amount. The displayed components sum to $4.221bn, with rounding explaining the headline. [R2]

That estimate is historically useful as an example of the transparency problem, not as a measured current total or an audited industry census. Extrapolating nondisclosers requires assumptions that are not interchangeable with the companies’ accounts. Likewise, a survey’s estimate of outstanding consumer balances cannot be relabeled annual corporate revenue.

The data behind the story

F24. Most of a widely cited historical estimate depended on nondisclosers

USD million

Explore the data table

← Scroll left or right to see all columns →

F24. Most of a widely cited historical estimate depended on nondisclosers (USD million)

Category	2017 estimated breakage
Full disclosers	381 USD million
Hybrid disclosers	755 USD million
Nondisclosers	3,085 USD million

T32. What would make a defensible national estimate

What a transparent gift-card disclosure would reveal

The best disclosure would connect the customer promise to the financial result. It would state face issued, cash collected, redemptions, cash returns, legal remittances, remaining face value, carrying liability and recognized breakage. It would explain whether loyalty rewards or promotional instruments are included and identify the portion caused by estimate revisions.

That level of reporting would help several readers at once. Customers could distinguish access failures from actual loss; campaign buyers could see whether rewards reached recipients; analysts could separate sales performance from estimate changes; and policymakers could assess how much value was recovered, rather than counting only legal rules on the books.

A voluntary donation or balance-recovery scheme can be useful, but it must not silently replace the recipient’s rights or a legal duty to remit funds. The defensible question is not whether forgotten balances look available; it is who has authority to direct them and whether the intended beneficiary can still recover value.

T33. Questions for issuers, auditors and reward platforms

Checking widely circulated breakage claims

Widely repeated claims about unused gift card money often combine different years, products and accounting measures. The comparison below separates the reported evidence from interpretations that change its meaning: a survey balance is not annual revenue, an odds ratio is not a probability multiplier, and association does not establish intent.

T34. Claims that survive, change or fail verification

[U01] [U02] [U03] [U04] [C01] [C06] [C08] [A01] [A03] [A09] [A10] [A11] [L01] [L02] [L03] [L06]

Historical billions are not one continuous series

CEB’s original releases projected $1.7bn of spillage in 2012, about 1% of spend in 2013 and less than 1% in 2015. An analysis by the New Jersey Office of Legislative Services describes a definition involving expiry, fees and business closure, and anticipates temporary suppression from the five-year expiry floor. The series therefore cannot be treated as directly comparable to current proportional accounting recognition. [U06] [U17] [U07] [U08]

The approximately $21bn reported in the 2022 consumer survey and $27bn in the 2024 survey estimate outstanding unused balances. Those are stocks of value held at the time of each survey. The earlier spillage estimates concern value not spent under a different definition. Combining the two measures into a single trend would confuse money still available to consumers with money treated as lost or unspent. [U02] [U04] [U08]

T35. A guide to the conflicting national numbers

[U06] [U17] [U07] [U02] [U04]

Methods and limitations

Core research cutoff: 15 September 2026; the Statista additions were reviewed on 16 September 2026. The study uses company filings, original survey releases, regulator materials, full academic papers and attributed secondary data tables. Statista is credited alongside the underlying survey or reporting organization, and republications of the same study are not counted as independent evidence. Publication and access dates are kept separate from observation dates; source fiscal years retain their original labels.

The company sample is purposive, selected for relevant public disclosures, and is not statistically representative. Dollar amounts are nominal and retain their reported currencies. No foreign exchange conversion or inflation adjustment is applied. Missing disclosures remain missing; immaterial is a qualitative statement, not zero. Derived changes use unrounded source values when available.

The charts distinguish observed disclosures, survey estimates, historical research and illustrative scenarios in their captions. Survey means are conditional on their stated populations. Multiple-response categories are not additive. No synthetic customer records are presented as evidence. Model examples teach mechanisms and sensitivity; they do not supply a forecast for any real issuer.

Academic findings are drawn from the full texts linked in the source register, with the February 2026 working-paper version identified explicitly. Every interactive figure includes an expandable data table. Percentages calculated here retain their stated denominators; approximate source values remain approximate. Reproducing arithmetic does not independently validate a survey, company estimate or statistical model.

The unanswered empirical questions are explicit: a current comprehensive national ultimate-breakage total; comparable category-wide breakage rates; causal evidence for a universal small-balance threshold; an issuer-level open-loop entitlement census; and a broadly comparable digital-versus-physical permanent-loss estimate. None is filled with a guessed number.

T36. Evidence labels used throughout