When to fire your payment processor: 7 signals every operator should know
- Switching processors is expensive. Staying with the wrong one is more expensive. The discipline is knowing which signals cross the line from inconvenience to structural problem.
- The seven signals: unexplained freezes, declining auth rate, rising chargeback escalation, reserve growth, account manager churn, policy drift, and roadmap divergence.
- Any two of the seven crossing threshold simultaneously means start sourcing a replacement now. Three means cutover within 90 days.
On this page
Operators stay with bad processors for the same reason tenants stay in bad apartments: the cost of moving is visible, the cost of staying is invisible. But in payment ops, the invisible cost compounds. Every month on the wrong processor is another month of suppressed conversion, rising reserve holds, and executive attention absorbed by account drama that should not exist. This is the framework we use to decide when to fire a processor — with thresholds, not vibes.
1. Signal one: unexplained freezes or fund holds
A freeze event is the loudest signal, and most operators correctly recognize it. The nuance: one freeze event can be a bad month or a misaligned risk model. Two freeze events on different brands in the same calendar year is a pattern. Three events in two years is the processor telling you the relationship is not working.
The threshold: any single freeze that holds more than 20% of a brand's 30-day volume for more than 45 days, or any two freezes of any magnitude on different brands within 12 months. Either condition triggers the signal. See our notes on Stripe freezes and vertical-specific freeze patterns.
The reason to take freezes seriously as a systemic signal: freezes are downstream of the processor's risk model disagreeing with your business model. The disagreement does not resolve with a call to the account manager. It resolves by moving to a processor whose risk model is compatible with the verticals you operate in.
2. Signal two: declining authorization rate
Authorization rate is the percentage of attempted card charges that get approved by the issuer. Healthy auth rate is 88–94% for DTC depending on vertical and fraud profile. Anything below 85% is a material conversion problem.
The threshold: a trailing 90-day auth rate that dropped more than 2.5 percentage points from the 12-month baseline and has not recovered. The drop signals either a processor-side routing problem (they are sending your traffic through a less-optimal acquiring bank) or an issuer-side decline problem (issuers are flagging the processor's descriptor, not your brand).
Before firing on this signal, have two conversations: one with your account manager asking for routing optimization, one with your own fraud vendor asking if their rules have tightened. If both come back clean and the auth rate does not recover within 60 days, the signal stands. Every percentage point of auth rate on $1M/month of volume is $10k of monthly top-line revenue — $120k a year. Do not let this one slide.
3. Signal three: rising chargeback escalation
Good processors escalate chargebacks proportionally to your rate of disputes. Bad processors escalate disproportionately — putting you into an early warning program on a normal ratio, threatening threshold breach for a single bad month, requiring chargeback reduction plans at ratios below the network threshold.
The threshold: any written communication from the processor threatening program enrollment or penalty at a chargeback ratio below 0.9%. The Visa and Mastercard network thresholds start at 0.9–1.0%; processors who pre-escalate below that are either running their own stricter policy or signaling they want you off the platform.
A pre-escalation letter is the processor telling you the risk department has flagged your account. The friendly version of the conversation does not exist anymore. Even if you clear the current issue, the next chargeback spike will trigger closure. See our chargeback dispute playbook for interim damage control, but start sourcing replacement processing the week the letter arrives.
4. Signal four: reserve growth
Some reserves are normal. Rolling reserves at 2–5% of 30-day volume are standard for many verticals. What is not normal: upward reserve revisions mid-contract without a corresponding event justifying them.
The threshold: any reserve increase of 25% or more on top of the baseline, applied without an identifiable trigger event (like a chargeback ratio breach or a large refund cluster). A silent reserve raise is the processor de-risking their exposure at your expense. Once they have started, they will continue.
Our reserve calculator shows what a reasonable reserve looks like for your profile; if your current reserve is more than 1.5× that number and growing, the signal is live. Reserve holds in high-risk verticals covers the vertical-specific baselines.
5. Signal five: account manager churn
This signal sounds soft but is remarkably predictive. If you have had three different account managers in 12 months, the processor is restructuring internally or has deprioritized your segment. Either way, institutional memory about your account is zero. Every policy conversation starts from scratch. Every escalation is a cold call.
The threshold: three account manager changes in 12 months, or any single change where the new rep does not proactively introduce themselves within 30 days. In a healthy processor relationship, a new AM reaches out on day 1. In a deteriorating one, you discover the change by email bounce.
The corollary: if you have had the same AM for 3+ years, that is a protective factor. Do not fire a processor with a strong AM on the other four signals alone. A skilled AM can route around a lot of structural problems. Without one, the structural problems become your problems.
6. Signal six: policy drift
Policy drift is when the processor's acceptable-use policy tightens in ways that catch verticals you operate in. The processor may have accepted peptide sales in 2023 and sent a policy update in 2026 reclassifying them as high-risk. Your existing account is grandfathered — until it is not.
The threshold: any policy amendment that names a vertical you operate in, with or without immediate enforcement. Even grandfathered acceptance becomes enforcement during the processor's next risk review. Once the policy exists, the enforcement is a question of when, not if.
Our when to move off Square piece covers the Square-specific pattern; the same logic applies to any processor whose published AUP has drifted against your portfolio.
7. Signal seven: roadmap divergence
The softest but most important long-term signal: the processor's product roadmap is not building the capabilities you need. If you need multi-currency settlement, 3DS-2 flexibility, multi-acquirer routing, or marketplace-style splits — and the processor's roadmap for the next 18 months is about incremental UX polish on their single-account flow — you are about to outgrow them.
The threshold: any single product capability you need that will not be delivered in the next 12 months, with no reasonable workaround. The signal is loudest on multi-brand operators because their needs diverge from single-brand operators faster. Stripe is a great tool for brands 1–3; the features you need at brand 8 are not on Stripe's roadmap and will not be. That is not a criticism of Stripe — it is a mismatch of tool to scale.
8. The two-signal rule
Any single signal can be a bad month or a bad quarter. Two signals crossing threshold simultaneously is a pattern. Three signals is an emergency.
Two signals: start interviewing replacement processors within 30 days. Have a shortlist of 2–3 acquirers by day 60. Underwrite with one of them by day 90. Do not commit until you see how the first two signals evolve — they might resolve. But do not be unprepared either.
Three signals: cutover within 90 days regardless of whether the original signals resolve. Three crossed thresholds means the processor relationship has structural problems that will surface in new ways even if the current ones get patched. See our 30-day cutover plan for the sequencing.
9. The signals that are not signals
For completeness, a few things that feel urgent but are not fire-the-processor triggers: a single rate increase during contract renewal (negotiate it), a single support ticket mishandled (escalate it), a one-time delayed payout due to banking holiday (confirm it was the holiday), a new 3DS requirement on a specific card brand (roll with it), a single-brand chargeback spike caused by a known bad SKU (fix the SKU).
These are operational friction. They are not structural incompatibility. Leaving a processor over them means incurring switching cost without solving a problem. Save the exit leverage for the signals that are actually structural.
10. The switching decision, one last time
The cost to switch a 12-brand portfolio: $25–35k in engineering and project management, 30–60 days of elevated operational attention, and ~120 days of parallel chargeback handling. The cost to stay with a processor showing 3+ signals: typically $150–400k/year in suppressed conversion, excess reserves, and crisis-mode ops time — plus a non-zero probability of a portfolio-scale freeze event that costs 10× the switching cost on its own.
Math favors switching in every case where three signals are live. Math often favors switching where two are live. Math rarely favors switching where only one is live. Run your own portfolio through the seven-signal checklist this quarter and be honest with the answers. If you want a second opinion on whether the signals you are seeing cross threshold, send the intake or see what switching actually costs.