Bread Financial Holdings, Inc. BFH

Key data

The business

Bread Financial, the old Alliance Data Systems, is a lender dressed as a checkout offer. It issues the private-label and co-brand credit cards that retailers hand you at the register, runs a small buy-now-pay-later product and takes online savings deposits to fund the loans. The money is made the plain way a bank makes it: charge cardholders interest and fees, pay the retail partner a cut, subtract what people fail to pay back, keep the rest. The moat is switching cost and scale on the partner side, retailers do not re-plumb a card program lightly, but the borrower is lower-income and rate-sensitive, so the whole thing lives or dies on one number, how much of the loan book goes bad. That is the pane of glass to keep clean. This company has been burned before by losing a large partner, which is why every renewal matters as much as any credit chart.

The numbers

The story is a recovery, and the shape of it is violent. Look at full-year profit: it swings because credit costs swing.

Profit nearly doubled from 2024 to 2025 as loss provisions came down off a fear-driven peak, and the first half of 2026 is already tracking ahead of the first half of last year. The recent quarters carry the same signal.

Every one of the last four quarters landed far above what analysts modeled, $3.55 against $2.74 last quarter, $4.18 against $3.00 the quarter before. Beats that size are not luck, they mean the Street is still pricing a recession-level loss rate that has not shown up. The precise monthly charge-off figures sit outside this run's data, but the earnings recovery and the run of beats confirm credit costs fell, and the monthly loss rate remains the thing to refresh.

On the math, this is a compounder hiding behind a cyclical: tangible net worth is roughly $65 a share and earnings power is near $13, a return on tangible equity around 20%. The stock throws off about a 12% earnings yield, and management is spending it, $313M of buybacks in 2025 and another $153M in the first quarter of 2026, enough to retire 5% to 7% of the shares a year. Diluted shares have fallen from about 47.6M to roughly 41M. So even flat earnings compound per share at high single digits, and a re-rate from 1.7x toward the 2.0x rivals fetch would add nearly a fifth on top. The variant view is simple: the market is paying 9 times earnings for a subprime-store-card lender it expects to break, and credit already normalized. The print that settles it is the next few monthly master-trust loss files.

Management

CEO Ralph Andretta runs the ledger the way the numbers suggest, redeeming debt and buying stock hard while the shares sit below tangible-book multiples of peers. The tell to weigh against that: insiders have been net sellers, five sales worth about $2.5M against a single $60K purchase over the last year, including Andretta's own $1.1M in May and CFO Praniti Lakhwara's $297K in late July, all with plan status not disclosed. Selling into a stock that has doubled off its low is ordinary, not alarming, but it is not the buying you would want alongside the buyback. The guidance record cuts the other way: four straight quarters of conservative outlooks blown through means the company is either sandbagging or genuinely surprised by its own credit, and both favor the owner.

How it fails or surprises you

Credit re-accelerates. The borrower here is the first to fall behind when jobs soften. If net charge-offs climb back toward and past 8% and delinquencies rise with them, provisions eat the earnings beat overnight, the way 2022 and 2024 gutted profit. Watch the monthly master-trust loss rate breaking higher for two consecutive months.

The earnings are not stable. The clearest argument against this memo is its own table: profit went $801M, $223M, $718M, $277M, $518M in five years. Paying 9 times a good credit year can be a trap if you are buying the top of a loss cycle. The renewal or loss of a large retail partner would swing the loan book the same violent way.

Normalization plus buyback compounding (right tail). If losses simply hold in the 7s, the $13 of earnings power is real, the multiple drifts from 1.7x toward peer 2.0x tangible book, and the share count keeps dropping 5%-plus a year. The market is still paying recession odds. Another quarter of low-7s losses and a stable delinquency print would start to force the re-rate.

Closing thoughts

The monthly credit data decides this, visible within a quarter or two. If loss rates hold where the earnings beats imply they already are, then 9 times earnings on a 20% return on tangible equity is too cheap and the value closes as the multiple lifts and the buyback shrinks the count. If losses break higher, provisions swallow the profit and the cheapness was a warning. On the weight of the evidence, the earnings have recovered and the beats keep landing, so the upside tail is the fatter one. What is genuinely at risk is a recession that drives losses to 10%-plus and halves earnings the way 2024 did; what it is worth if credit behaves is a re-rate stacked on high-single-digit per-share compounding.

The bet is still that American shoppers keep putting everyday purchases on store credit cards and keep paying those balances back with interest, and that Bread's retail partners keep renewing their deals. What breaks it is the pair to watch, the net charge-off rate turning up while delinquencies rise alongside it. If those two climb together for two straight months, the read here is wrong.