CASE STUDY 01/CHAPTER 2 OF 3 · BUSINESS CASE AND MIGRATION

Moving the Apple Card to a new issuer

On January 7, 2026, Apple named JPMorgan Chase as the replacing issuer for roughly 12 million cardholders and more than $20 billion in balances. That is the only announced fact.

Everything else here is the migration I would recommend if I were consulting both sides, built as a 100,000-account re-underwriting simulation across a 36-month horizon.

AT A GLANCE
ROLE
Sole author, end to end
Built the simulation, set the routing and reward assumptions, ran the stress tests, and wrote every word here.
DISCIPLINE
Product strategy, business casing, growth and lifecycle analytics
Business case and financial modeling, unit economics, retention and lifecycle analysis, LTV to CAC, scenario and sensitivity testing, partner and commercial negotiation
BASIS
100,000 synthetic accounts, 10 migration waves over 18 months, 36-month projection
STATUS
Proposed concept, not an announced plan
LAST REVIEWED
August 14, 2026
Model unchanged since the June 25, 2026 assumptions note.
ISSUER-SIDE BUSINESS CASE

The business case

Proposed concept on synthetic account data. The issuer change is announced, the migration design is mine.
What each side earns, per account and in sample
$100.34M
What Chase earns from the sample over three years, counted in today’s dollars
$85.74M
What Apple earns on hardware and services over the same three years, same sample
$1,003
What Chase earns per account, the unit every total here is built from
$857
What Apple earns per account away from the card, which loses money on its own
Growth
10:1
Every dollar spent winning a customer returns about ten, against three as the healthy line
59.26%
Of cardholders routed to Chase Freedom Rise who go on to graduate to the Apple Card
Risk
95%
Of profit still earned if loan losses jump three percentage points
5 factors
Weighed on every account, so how someone actually pays can outrank their credit score
Loyalty
54.0%
Of accounts still active after three years, across all three outcomes
$965,373
What the sample gains if retention rises one point, say 54% to 55%

Every figure is an output of my own simulation model, not a reported result. I state these per account and as a sample total rather than scaled to the full portfolio. Multiplying by 120 would reach the announced 12 million cardholders and produce a number in the low tens of billions, but these are synthetic accounts I constructed rather than a random draw from the real portfolio, so scaling them enlarges the construction rather than estimating the portfolio. The announcement also gives more than $20 billion in balances, implying roughly $1,667 per account, and reconciling the model against that figure is the check I would run before quoting any portfolio total.

What the model says, in four points
01The value is set at re-underwriting, before any of the launch workMigrations are usually run as an announcement, with the effort going into emails and support scripts. By then the money is already decided, by who re-qualifies for the Apple Card (2028), who routes to Chase Freedom Rise, and who is turned down. Goldman never got to make those calls. It built its customer base by approving applicants one at a time as they came in, rather than choosing who to keep, and the losses followed. Handing the same accounts across unchanged would repeat the mistake, so the re-underwriting is the point of the migration rather than an obstacle in it.
02Chase loses money for four months, and that is normal rather than a warning signOn Chase’s side of the deal the portfolio is underwater for four months and bottoms out at $1.69M. That is mostly because banks must set money aside for expected defaults the moment an account opens, long before anyone actually misses a payment. Chase turns positive in month 5. Apple’s position is separate, and is set out alongside it later.
03Behavior drives the declines, not the credit score58% of the people turned down have a credit score at or above the old approval line. They are turned down for how much of their credit they have already drawn and for missed payments, which is what makes the decision explainable to the person receiving it.
04Who pays for Daily Cash changes the deal, not the credit decisionsChanging how much of Daily Cash Apple funds moves money between the two companies and nothing else. Who stays, who gets which card, and who is turned down are identical either way, so the two firms can negotiate terms without reopening the credit policy.

Each point is worked through in the sections that follow, with definitions in the glossary.

01

What is known, and what I am proposing

One fact is public. Everything else on this page is a design I am putting forward, and the line between the two matters more than any single number.

ANNOUNCED
The issuer is changing

On January 7, 2026, Apple named JPMorgan Chase as the replacing issuer, covering roughly 12 million cardholders and more than $20 billion in balances. Goldman Sachs exits. That is the extent of what has been made public.

PROPOSED, BY ME
Everything about how

The three-outcome structure, the wave schedule, the risk score, the reward split, and every figure here. None of it is an Apple or Chase plan. It is what I would recommend if I were advising both sides.

Why a re-underwriting, and not a transferA portfolio sale can move accounts as they are. But the reason this portfolio is changing hands is that its credit performance disappointed the outgoing issuer, so handing the same portfolio over unchanged would hand over the same problem. The migration is worth modeling precisely because it is the one moment a new issuer can reprice risk it did not write.
02

Three outcomes, ten waves

Every account is re-reviewed and lands in one of three places. Two are products. The third is not a product at all.

OUTCOME 1 · RE-QUALIFIES
Apple Card (2028)

Passes the new credit review and moves onto the relaunched card. Premium benefits, no annual fee, and no sign-up bonus, since welcome offers have always been for new applicants and everyone here already holds the card. The rate each cardmember pays is set by their credit grade.

68.4% of accounts routed here
63.8% of those retained at 36 months
OUTCOME 2 · ROUTED
Chase Freedom Rise

Does not clear the bar for the Apple Card (2028), so routes here instead of losing access entirely. Builds credit on it, with a documented path to graduate onto a new Apple Card (2028) account later. Declined applicants who reapply arrive here too, not at the Apple Card, unless they clear the top range outright.

21.9% of accounts routed here
47.1% of those retained at 36 months
OUTCOME 3 · DECLINED
Responsible decline

Clears neither bar, so no card is issued. Rather than an abrupt cutoff, the account follows a managed pathway to closure, with a written notice naming the reason and what would have to change to qualify. This is the outcome that has to be defensible in writing, and it is the subject of its own section. These are still future customers, and consumer finance is the one place Apple cannot afford a bad story, so a migration that reads as mass account closures writes that story for it.

9.7% of accounts, no card issued

How the rollout is stagedThe rollout runs in 10 waves across 18 months rather than all at once, which keeps the operational load and the support volume survivable, and gives the credit policy nine chances to be corrected before the last wave moves. Blended across all three outcomes, 54.0% of the original accounts are still active at month 36.

Why three outcomes and not twoChase Freedom Rise as the middle path is what I would argue hardest for. The easy version of this migration has two outcomes, approve or decline, and it is worse for everyone: Chase loses accounts it could have earned later, Apple loses customers from the ecosystem entirely, and the customer loses access at the exact moment they are least able to replace it. A middle path costs something to stand up and it is the difference between a re-underwriting and a purge.

What it means for the cardholderIt also matters for the customer in a way the economics do not capture. Closing an account ends the credit history attached to it, and length of history is one of the things a score rewards. Routing someone to Chase Freedom Rise keeps that relationship open and reporting. The smaller credit line will cost them something, but a lower limit is a scratch next to losing the account outright.

03

Who the portfolio actually is

What the groups are worthFour behavioral groups, and they are worth very different amounts to the issuer. The credit grade says how likely someone is to repay. The segment says how they actually use the card, and that is what decides whether the account earns anything.

Share of accountsTO SCALE
Daily Transactor43%
Ecosystem Maximizer27%
Preferred Maximizer18%
Installment Revolver12%
Share of portfolio value
Ecosystem Maximizer66%
Everyone else34%

The inversion is the finding. The largest segment by headcount is not the largest by value, and the volume block earns modest interchange spread thinly. Retention spend, cross-sell, and wave order all follow the second chart rather than the first.

Both charts are outputs of my own simulation across 100,000 synthetic accounts, not reported results.

The segment worth keepingThe Installment Revolver is the group a business case usually deletes, and I would keep it. On the card by itself it loses money, and that is the point. These are people who used the card to pay for an iPhone over time and then never take it out again. Chase earns almost nothing from them, and Apple sold the phone because of them. Look only at what the card earns and you cut the thing that moved the hardware, which is why the two sides have to be read together. It is also the group with the most room to improve. They already carry the card. They just have no reason to use it anywhere else, so a rate worth reaching for turns a financing account into a spending one, and that is the cheapest new spending the issuer can get.

04

The economics, and the shape of the curve

The portfolio loses money before it makes any, and the reason is worth understanding, because a reader who sees only the first four months would kill this deal.

What sits in front of the payback
01Loss reserves. Money booked up front against defaults the issuer expects later, before a single one occurs. On a freshly underwritten portfolio this is the largest early charge, and it is the main reason the curve starts underwater.
02Onboarding and marketing. The cost of standing the portfolio up across ten waves.
03Chase's share of the early cash back. Apple pays most of the Daily Cash and Chase pays the rest. In the first few months Chase's share runs near 27% rather than the 25% it settles at later, because the earliest waves are already spending before the agreed split takes full effect.
$1.69M
Deepest cumulative loss, reached in month 2
Month 5
Cumulative contribution first crosses zero
What this figure is not

The portfolio total and the month 5 payback are what Chase earns, and nothing else. What Apple spends on rewards and what it makes back on hardware and services are counted separately, and neither one changes this number. With two companies splitting the work, the easiest mistake is reading one company's profit as the value of the whole deal.

05

Re-underwriting to risk, on two dials

Goldman’s losses came from credit that did not match risk. There are only two dials an issuer controls here: the price of credit, and how much of it someone gets. The first chart is price, the second is exposure.

Dial one, price: the interest rate by credit grade
Super-prime18.49%
Prime22.99%
Near-prime26.49%
Subprime29.74%

Riskier grades pay a higher rate, and every grade still earns money on its own. Bars are scaled to the top rate.

Dial two, how much credit: the total of every credit limit, before and after
Goldman Sachs$1.74B
Chase$1.87B
+13%
Low-utilization super-prime lines rise, against their own starting lines
about 80%
High-utilization lines cut
54% lower
Lifetime charge-offs
76%
Of the riskiest borrowers, the ones already near their limits, end up on a capped Chase Freedom Rise line or no card at all, so the credit that caused the losses cannot be handed back out at the same size.

Chase hands out more credit overall and still loses less, because the limits move toward the people who can carry them. The two bars share one dollar scale so they can be compared directly. The percentages are not on that scale, because each one measures a different group against a different starting point, so they are written out rather than drawn. Anyone whose limit drops more than 20% gets a written notice explaining why.

Both charts are outputs of my own simulation across 100,000 synthetic accounts, not reported results. Bars are scaled within each chart, and the two charts are not on a shared scale.

06

The declines, and why behavior beats the score

This is the outcome that has to survive scrutiny, so it is the one I built most carefully. The finding is counterintuitive, because most declined accounts have a perfectly acceptable credit score.

58%
Of declined accounts score at or above the old cutoff. They are turned down on utilization and delinquency, which the Migration Risk Engine weighs heavily, not on the score itself.
621 FICO Score 8 · 51% debt to income
The declined group's averages. Half of monthly income already goes to servicing debt, which is the number that makes further credit unaffordable regardless of what the score says.

The 58% — the number I would lead withThat 58% is the number I would lead with in a room, and not because it flatters the model. It is the number that proves the policy is not a score cutoff with extra steps. If a majority of declines clear the old bar, then the old bar was measuring the wrong thing, which is a plausible account of how the incumbent ended up with a portfolio it did not want. A decline is also the one decision here that a customer experiences as a judgment about them, so it should be the decision with the most reasoning behind it.

07

Who pays for the cash back, and why it changes nothing else

Every dollar of Daily Cash a cardholder earns has to come from somewhere, and Apple and Chase split the bill. There is no convention to fall back on here. As I understand the common airline or hotel co-brand, the issuer funds rewards by buying the partner’s miles or points, which makes the reward cost revenue on the partner’s side. Daily Cash is cash, so Apple has no currency to sell and that mechanism does not exist. Someone writes a real check, and who writes it is negotiated rather than inherited. I ran two versions of that split, and the interesting part is what stays exactly the same in both.

Who funds Daily CashTO SCALE
A tougher deal for Chase, tested onlyApple pays 55.8%, Chase pays 44.2%$73.74M to Chase
55.8%
The split I would push for, and the one every figure here usesApple pays 75%, Chase pays 25%$100.34M to Chase
75%
Paid by Apple
Paid by Chase
Nothing about the customer changes

Who gets approved, who gets routed to Chase Freedom Rise, who gets declined, how many stay, what Apple earns on hardware: all identical in both versions. Changing the split only moves money between Apple and Chase, and it moves a lot of it, a $26.6M swing on identical credit decisions.

Asking Apple to cover 75% is an aggressive opening, and I would say so in the room rather than pretend otherwise. What makes it arguable is that Apple is not trying to earn on the card in the first place, so the funding it takes on is acquisition spend against hardware and services rather than a loss. What makes it negotiable is that nothing in the credit policy depends on the answer, so the two companies can settle the underwriting on its merits and then move the money between them without reopening any of it.

Worth being clear about which one the page runs on. Every figure in this chapter is calculated at 75% Apple funded. The tougher split is there to test the argument, not as an alternative set of results. Under it Chase pays more of the cash back, so its own numbers come down while nothing about the customer decisions moves.

08

The money that never shows up on the card

What the card earns is not the whole story. What the card earns already counts the two things a card normally earns, the fee on every purchase and the interest on balances people carry. The savings deposits the card brings in are tracked separately, so they sit outside that figure rather than inside it. And it counts nothing at all of what Apple makes on the iPhones and subscriptions the card helps sell.

Why the loss is the pointApple’s per-account loss is the most misreadable number in this model, and the one I would put in front of a skeptic first. It looks like a failing product. However, it is a loss leader, which is one of the oldest plays in retail. Sell one thing below cost to move the things that carry the margin. Apple is doing that with a credit card instead of a grocery item, and services margins run well above hardware, so subscription revenue from a retained customer covers it fast.

The same trick in reverseThe deposit line is the same trick in reverse. As a spread it is noise next to what the card earns. As a wedge it is the only line here that makes Chase someone’s bank rather than one of their cards, and everything Chase could sell that customer next, checking, a mortgage, a travel card, sits outside this model entirely.

What I cannot verifyWhat I cannot do is check either one. The Apple and Goldman revenue split was never disclosed, so the Apple side of this model is estimated rather than reported, and that is the honest limit on it. What I would defend is the shape. Both numbers look like failures in the place people look and pay for themselves in the place they do not.

09

What breaks it

A business case that only shows the good case is a pitch. So I broke the model on purpose to see what survives.

What a stress test asksThe thing most likely to break a card portfolio is people not paying it back. A stress test asks what happens if that gets much worse than expected, and the standard way to run one is to pick a number a regulator would recognize rather than one that flatters the result. I raised charge-offs, the share of balances written off as unrecoverable, by three percentage points across every credit grade. That is the order of magnitude I would treat as a downturn case, and it lands well above the 3.2% Chase currently guides for its own card business.

The reason to run it is that every figure on this page rests on losses behaving as modeled. If a modest worsening wiped out the return, the case would only hold in good weather, and nobody underwrites for good weather.

There is a second reason, and it follows from the case for re-underwriting rather than assuming the portfolio as it stands. The test is really asking whether the migrated portfolio holds up under pressure the legacy one did not, and if it does not, the routing thresholds are wrong rather than the stress being unfair.

Loan losses raised three percentage points across every credit grade
$94.9Mstill earned, 95% of the base case
$5.4Mlost to the shock

Drawn to scale against the $100.34M base case. The portfolio stays comfortably positive, which is the result the underwriting is supposed to produce. The point of declining accounts on behavior rather than score is that the surviving portfolio is not fragile. Losses are the risk I can test, though, and not the one I would worry about most. If people leave faster than the model expects, each percentage point of retention costs about $1M, so being five points off would cost roughly as much as this entire loss shock. Loan losses are the risk with real history behind them. Retention is the one I am guessing at.

10

The recommendation

Three things decide whether this migration works, and they are the three the headline figures are grouped under: who carries risk, who stays, and who spends.

Risk, decide it once and decide it early

Review every account again rather than moving them across as they are, and give people three possible outcomes instead of two. Move them in ten waves over 18 months, so the rules can be corrected nine times before the last wave goes. Chase does not want to inherit the cardholders who ran their limits up and stopped making even the minimum payment, because that cohort is why the Apple Card wrote off a larger share of its balances than comparable issuers.

Base the decision on how someone actually pays rather than their credit score alone, and be able to explain every rejection in one sentence the person would recognize as fair.

Loyalty, keep the people you would otherwise lose

Chase Freedom Rise is what keeps people who would otherwise walk away for good, and it costs very little to offer. 59.26% of the people sent there end up holding the Apple Card anyway.

Why retention compoundsThe customers worth keeping are already tied to Apple through their hardware and subscriptions, and that tether is what makes them reachable for Chase at all. Every account retained is an account Chase can sell its other products to, which is what the deposit argument rests on. Chase Freedom Rise makes that concrete. As the product stands today it requires a Chase checking account, so anyone routed there who does not already bank with Chase opens one as a condition of receiving the card. The product chapter proposes widening that, letting the Apple Card Savings account these cardholders already hold satisfy it instead, which trades a new checking relationship for a lower barrier to accepting the offer.

Growth, the part that does not happen at launch

Keep marketing the card after the migration is done. Better approvals and better rewards decide who holds it and what they earn, but neither gets it out of the drawer. Chase only collects the merchant fee if people actually spend, and the group who only finance hardware are the easiest ones to win over, because they already carry the card.

What the migration leaves behindThe migration also leaves something behind that is worth more than the portfolio itself. Re-underwriting 12 million existing accounts produces a record of which behaviors actually preceded losses on this specific portfolio, and that record is what should set the bar for everyone who applies afterward. That is the one chance to calibrate before opening the doors again, which is what the acquisition offer in chapter 1 is built on top of.

How I would negotiate itGo into the negotiation asking Apple to cover 75% of the cash back, but treat that as a price rather than a requirement. None of the credit rules depend on it, so it can be traded away without reopening anything that matters.

WHERE I PUSHED BACK ON MYSELF

The limits of this chapter

The accounts are synthetic. Nobody outside Apple and Chase has the real portfolio, so the 100,000 accounts are simulated against published behavior and rate data. That means the mechanics here are testable and the specific dollar outputs are not verifiable against reality. Treat the structure as the contribution and the figures as what that structure implies under my assumptions.

One announced fact, one modeled everything. The issuer change is public. The product lineup, the routing rules, and the reward split are all mine. A reader who takes this as reporting on Apple's plans has misread it, which is why the distinction sits in the first section rather than a footnote.

Response to an incentive is unmeasured. Nothing here prices what an offer buys, in retention or in acquisition, because I could not source a figure for it and the record does not contain one. That is why the acquisition offer in chapter 1 is specified as a structure with its match rate flagged for testing rather than costed here. Sizing either would need a live holdout, which is the first thing I would instrument.

The hardware baselines are stale, in the conservative direction. Apple raised prices on several lines after the model was built. I left the pre-increase baselines in and reported the sensitivity rather than quietly restating the headline, because a figure that moves when you refresh an input should be labeled, not improved. Raising the baseline 10% is the sensitivity I ran.

Retention is modeled, not observed. The 63.8% and 47.1% figures come from my own churn assumptions by product, so if those hazards are wrong the retention economics move with them.

The calibration argument cannot be tested on this data. Re-underwriting a real portfolio would teach an issuer which behaviors preceded real losses, and that is the case I make for it. This model cannot demonstrate it, because simulated accounts default according to the hazards I assigned them, so scoring them against their own performance recovers my assumptions rather than anything about credit. The argument is a reason to run the exercise, not evidence from having run it.

What Apple earns is estimated, not disclosed. The Apple and Goldman revenue split was never published in either company’s filings, so the Apple side of this model is built from what the card plausibly moves in hardware and services rather than from anything reported. That is the weakest-sourced half of the two-ledger argument, and the half I would most want a practitioner to challenge.

METHOD AND MODEL

How the numbers were produced

REFERENCE

Glossary

Independent professional-development project. The migration design is a proposed concept built on synthetic account data, not an announced plan, and it is an Apple-inspired concept rather than an official Apple product. Not affiliated with, authorized by, or endorsed by Apple, JPMorgan Chase, Goldman Sachs, Visa, or Mastercard. The public facts relied on are listed in the sources. Every modeled figure is an output of my own simulation, which is not published, and none of it is legal, financial, or investment advice.

PRODUCT STRATEGY, PRODUCT MANAGEMENT, PRODUCT MARKETINGGROWTH MARKETING, LIFECYCLE MARKETING, BUSINESS DEVELOPMENT