Keep, downgrade, or cancel: the framework

The fee posted yesterday. Since then you have opened the issuer's app, a forum thread, and a spreadsheet, and closed all three without deciding anything. The renewal window is ticking, the retention number is sitting somewhere in your phone, and every search result gives you a different answer dressed as certainty.

Strip away the noise and there are exactly three answers: keep the card, downgrade it to a cheaper product from the same issuer, or cancel it. The right sequence is equally short. Audit the card's value first, because a decision made before the audit is a guess wearing a lab coat. Try retention before you exit, because a good offer beats every other branch. And know that for most people whose value math fails, the downgrade door is the correct one, since it keeps your points alive, your credit history intact, and often a status floor you did not know you were standing on. Cancellation is the rare case, not the default.

We will run one card through all three doors so you can watch the logic move: a Chase Sapphire Reserve that just charged $795.

Before any door: the thirty-minute audit

The break-even method, compressed to three sentences. List the card's hard credits at face value, but only the ones you would have bought anyway; add the extra earn above a free 2% card on your actual twelve months of spending; add perks priced at what you would honestly have paid. Divide by the fee and read the ratio, treating 1.2 to 1.5 as a sensible comfort band rather than scripture. Everything below assumes you have that number, because the three doors answer different questions for different ratios.

Our example: Maya, who has held Sapphire Reserve for four years. Her audit says $795 fee against $550 of value she will actually collect, a 0.69x ratio. Here is how she built it: the $300 travel credit posts against flights she books anyway; the dining credit, $150 twice a year at a specific reservation collection, is a behavior change she has never once made, so it counts at zero; six solo Priority Pass visits at her honest $30 apiece is $180; the earn delta over her no-fee flat-rate card is about $70 on her spending. That is $550. The card fails her margin test, and the doors open.

The three doors at a glance

Keep: when the card still pays for itself

The keep door is for ratios at or above your comfort band. The math says the card earns its fee, your behavior backs the math, and the honest move is to pay without guilt and re-audit next year. A card that clears 1.4x on credits you actually redeem is not a problem in your wallet. It is a discount program you understand.

Two refinements belong inside this door. First, a comfortable ratio does not forbid a retention call; it changes your posture. A keeper with leverage can ask about offers anyway, politely, once a year, because the worst outcome is a no. Reports through 2026 describe Chase offering Sapphire Reserve cardholders statement credits with no spend requirement, commonly $100 to $250 and occasionally $400, right at renewal (travelupdate.com, February 3, 2026, community-reported; offers are account-specific and never guaranteed). The timing and script guide covers the call.

Second, keep is a yearly verdict, not a lifetime one. Maya's 0.69x could become a keeper ratio if Chase adds a credit she would use, or if her travel doubles. The door you choose this January does not lock.

One more line item belongs in any keeper's audit: authorized-user fees, which the 2025 refresh cycle repriced alongside everything else. Sapphire Reserve moved from $75 to $195 per additional cardholder, and Amex Platinum sits at $195 as well (TPG, January 6, 2025; Forbes, April 24, 2026). If your household's value depends on a partner holding the card, the AU fee is part of your true annual cost, and a household paying $795 plus $195 is really paying $990 for the package. Run the audit at the household total, not the headline fee; on the plus side, the AU fee at least buys lounge access on both cards now.

Downgrade: the branch most people should take

When the ratio fails, the instinct is to cancel. The better instinct is to ask what the account is still holding, because a credit card account is three things stacked together: a fee, a points wallet, and a line of credit with a birthday. A downgrade throws away only the fee.

The points wallet first. Close the only card that holds a transferable currency and the currency can die with it: American Express Membership Rewards forfeit if you hold no other MR-earning card, per Amex's own FAQ, and Chase's education pages say the same about Ultimate Rewards on a sole UR card. Downgrading inside the family keeps every point; the full anatomy of what survives a closure is its own story. Maya's 180,000 Ultimate Rewards balance survives a move to Sapphire Preferred or a Freedom card, which is reason enough on its own.

The line of credit second. A product change keeps the tradeline: same account, same open date, generally the same limit. There is no utilization spike and no age penalty, which makes downgrade the score-neutral exit; the full mechanics are worth reading before you assume closing is harmless. Maya's four-year-old account with a healthy limit keeps working for her score at a $95 fee instead of $795.

The status floor third, mostly for co-brand holders. Airline and hotel points live in the loyalty program, not the card, so they survive even a cancellation; what does not survive is the card-issued status, free-night certificates, and anniversary awards. A downgrade to a cheaper card in the same program usually preserves a status tier and sometimes a certificate, which a cancellation never does.

Picking the floor is a matching game: the floor must preserve what you hold. Ultimate Rewards: a Sapphire or Freedom product keeps the wallet alive, and Sapphire Preferred, refreshed June 10, 2026 with new earning categories and no fee change at $95, keeps partner-transfer rights too; the Chase downgrade guide maps the paths. Membership Rewards: any MR-earning card preserves the balance, though note that Amex suspended new Green Card applications in August 2026 and whether product changes into Green still work is not confirmed; verify with the issuer before counting on that rung. Program currencies like SkyMiles or Bonvoy: almost any card in the program keeps the earning spigot and often the status floor. Two timing rules govern every path: most issuers require the account to be at least twelve months old, and a product change never carries a new welcome bonus, at any issuer we track.

Maya's downgrade case: Sapphire Preferred at $95 preserves her UR balance, her transfer rights, her tradeline, and her portal earn, while dropping the credits she was not using anyway. Her 0.69x ratio at $795 becomes roughly break-even at $95. That is the shape of the right answer for most failed audits.

One nuance keeps the door honest: downgrades are one-way streets in practice. Moving Reserve to Preferred forfeits the Reserve benefit set for good, and if her travel bounces back next year, re-qualifying means a new application and a fresh Sapphire bonus clock under the 48-month rule that governs new-account bonuses. The door saves money reliably. It just does not swing both ways, which is exactly why the audit comes first and the downgrade comes last.

Cancel: the exit that is rarer than you think

Cancellation earns its place only when four conditions are true at once. The value math fails. No floor card fits, because the issuer's downgrade ladder does not hold what you need or you already hold the floor. Retention has been tried and came back empty or insulting. And you have no exposure: no points stranded in the issuer's currency, no certificate worth burning first, no loan application inside the next year that needs your limits. Most people who march toward cancellation fail at least one of these four, usually the first or the last.

When all four hold, run the exit in sequence. Park your points: transfer them out, cash them out, or move them to a keeper card in the same currency, and confirm the balance reads zero before anything else. Burn the perishables: free-night certificates and statement credits, because some certificates forfeit on closure per their terms and others expire on their own schedule regardless; Marriott free-night awards run roughly a year from issue whether or not the card survives. Then call: one retention attempt, honest and calm, with your audit number as the asking price. Finally, confirm the closure in writing and save it.

The written confirmation is more than a receipt. A month or two later, pull your credit report: the account should read closed at your request, in good standing, balance zero or amortizing as agreed. A clean closure needs no repair. A wrong one gets disputed while the phone transcript still exists.

One community pattern deserves its own flag: cardholders commonly describe a refund window for a just-posted annual fee, typically described as a few weeks and often cited around thirty days, when the account is closed. Refund windows are not published policy at most issuers, the exact terms vary, and the safe move is to ask directly at the moment of closure; not confirmed as a standard, so verify with the issuer before you rely on it.

Order of operations

Sequence errors cost more than decision errors. The rule that prevents nearly all of them: retention comes before any exit, every time. A retention offer can beat the keep math, the downgrade math, and the cancel math in one phone call, and you only learn what your account qualifies for by asking. A 0.69x card with a $400 statement credit offer temporarily becomes a 1.19x card, which is inside the gray zone where a one-year reprieve while you decide is a perfectly rational take. Maya's sequence: audit on day one, retention call in week one, downgrade decision in week two if the offer disappoints.

The second rule: never decide before the audit, and never decide in the app at 11 p.m. The fee posting is not a deadline; it is the starting gun. Leverage with the retention desk is highest right after the fee posts (NerdWallet, March 18, 2026), and your decision quality is highest after the math, so the calendar below is built around both.

A rough timing calendar for the renewal month:

Regrets worth avoiding

Two regrets dominate the mail we imagine receiving. The first: canceling a ten-year-old card over a $95 fee. The dollars are small and the age is enormous; a decade-old tradeline with a fat limit is doing quiet work for your score every month, and the deferred age effect lands years from now on a day you cannot yet name. Downgrade it, sock-drawer it with one autopay charge, do almost anything but close it. The second: canceling or downgrading past your transfer rights without noticing. Points that survive a downgrade to a no-fee card may lose their ability to move to airline partners, because what you can do with a currency depends on which cards you still hold; Chase's education material is explicit that partner transfers require an eligible card. A downgrade that keeps the balance but strands it in cash-out-only land is a worse trade than most people realize at the time.

A third regret is quieter: deciding by forum consensus. Community threads are where you learn what offers look like, and the retention landscape for 2026 is genuinely useful for calibration. But the threads are averages of other people's accounts. Yours has its own spend, its own tenure, and its own system-generated offers. Use the reports as a floor for expectations, not a forecast.

Your renewal-season workflow

Once a year, every fee card in your wallet gets the same thirty minutes: audit, retention call, door. That is the whole system, and this framework is the map for the door part. The tool that does the arithmetic part is the calculator, and it exists precisely so the ratio comes from your statements instead of your mood. If you hold several fee cards, stagger the audits across the calendar; each card's renewal date is its own decision point, and retention leverage peaks at each fee posting, not in bulk.

Run your card through our annual-fee calculator and get your keep-downgrade-cancel answer, then come back here with the number and pick the door that matches it.