Points are an economic promise
A rewards program encourages spending and retention by promising value after a qualifying activity. The issuer incurs a cost that depends on the program’s contractual terms, redemption choices and expected customer behavior. A point is not automatically a fixed-dollar expense, and the customer’s perceived value can differ from the issuer’s fulfillment cost. This distinction helps explain why programs can support attractive marketing while carrying substantial estimation risk.
American Express’s 2025 Form 10-K describes its Membership Rewards liability using expected redemption and cost assumptions. That is one issuer’s accounting disclosure, not a universal template for all card programs. Delta’s 2025 Form 10-K separately describes selling miles to American Express under their arrangements. Together, the filings illustrate that the issuer, the loyalty-program operator and the customer can occupy different sides of the same rewards transaction.
Earn, redeem and estimate
The program must measure points earned, points redeemed, outstanding obligations and expected future redemption. Breakage refers to rewards expected not to be redeemed, subject to the applicable accounting and contractual framework. It is an estimate of behavior, not an invitation to make redemption difficult or a guarantee that outstanding points have no value.
Changes in the redemption mix can alter cost even when the number of points redeemed is unchanged. Travel, statement credits and merchandise may have different fulfillment economics. Partner pricing can also change. A program manager therefore needs both a quantity view and a unit-cost view, with enough detail to distinguish customer behavior from commercial renegotiation and accounting-estimate changes.
A hypothetical liability sensitivity
Assume a program has one billion outstanding points, expects 80% eventually to be redeemed and estimates an average fulfillment cost of 0.8 cent per redeemed point. A simplified expected cost is $6.4 million: one billion multiplied by 80% multiplied by $0.008. This is a conceptual calculation, not a statement of the accounting required for every program or any issuer’s reported liability.
If expected redemption rises to 90%, the simplified cost becomes $7.2 million. If the redemption rate remains 80% but unit cost rises to 0.9 cent, the result is also $7.2 million. Both produce an $800,000 increase through different mechanisms. A financial review should identify which assumption changed, why it changed and whether the change also affects future customer economics.
Profitability requires the whole card relationship
Rewards expense should be assessed alongside interchange or discount revenue, interest, fees, credit losses, acquisition cost and servicing. A customer who pays in full can generate spending revenue while producing little interest income. A revolver can generate interest but also require funding and absorb credit losses. Treating all rewards spending as interchangeable obscures those differences.
Promotional bonuses deserve separate cohort analysis. A large acquisition award can create an initial loss that is justified only if the customer remains active long enough and produces sufficient contribution. Measure the behavior of customers attracted by the offer, rather than applying the average economics of a mature portfolio. Customers may also optimize across cards, shifting spend when a promotion ends.
Controls around the estimate
Recommended controls reconcile the points ledger to customer accounts and the accounting records, including reversals, refunds, account closures and partner adjustments. A merchant refund should not leave the system with inconsistent cash, purchase and rewards records. Manual adjustments need documented authority and an audit trail because small unit values can accumulate across a large portfolio.
Back-test redemption assumptions by , tenure and program type. An aggregate redemption rate can hide a new cohort behaving differently from older customers. Track estimate revisions and distinguish changes supported by observed behavior from changes driven primarily by a target financial result. The analysis should also consider how program changes alter future redemption, rather than assuming historic behavior remains stable after the customer proposition changes.
Tradeoffs in program design
Richer rewards may increase spend and retention, but they can attract customers whose behavior is less profitable than expected. More redemption options can improve customer value while increasing fulfillment cost or operational complexity. Restricting options can reduce short-term expense and weaken trust or retention. The useful decision is the net relationship outcome under transparent terms, not the smallest possible liability.
Partner concentration is another risk. A co-brand program may depend on a travel or retail partner for both customer acquisition and redemption value. Changes in partner economics, service quality or strategy can affect the issuer even when credit performance remains stable. Contracts, operational resilience and customer communications therefore belong in the rewards-risk review alongside the actuarial-style estimate of future redemption.
What would change the assessment
Evidence of durable incremental spending, retention and contribution would support a program’s economics. The conclusion should change when redemption accelerates, fulfillment costs rise, customer complaints reveal friction or acquisition cohorts fail to earn back their bonuses. A favorable accounting adjustment should be examined separately from underlying customer profitability.
The cited 2025 filings are dated examples of how major participants describe their obligations and commercial arrangements. They do not supply a live estimate for another issuer’s program. For a reader comparing card businesses, the key questions are what the reward promise costs to fulfill, how sensitive that estimate is and whether the customer relationship earns enough to support it after funding, losses and service.