The most expensive tickets on Amazon are free. Amazon Seller Support cases cost nothing to open, which is why almost nobody treats them as a skill — and why the same stranded-inventory issue that one seller clears in 48 hours costs another seller six weeks, a suppressed listing through a sales peak, and a five-figure hole nobody attributes to a support queue.
After managing hundreds of brands on Amazon, we’ve filed and worked tens of thousands of cases, and we’ll tell you something we look at in every account takeover: the case log. It’s more diagnostic than the ACOS. An account with 40 open cases, half of them auto-closed without resolution, three of them duplicates of each other, tells you exactly how the last team operated — and exactly how much resolvable margin is sitting in the queue.
The timing of this post isn’t random. The redesigned Seller Central is becoming the default worldwide through August, and the support surface is increasingly AI-first — Seller Assistant triages before a human ever reads your case. That changes the economics of case quality in one direction: a well-structured case gets resolved by automation in hours; a vague one loops for weeks. And every unresolved catalog, inventory, and account-health issue you’re carrying right now is an issue you’ll still be carrying in October, when queues slow, stakes rise, and nobody wants to touch a live listing.
Here are the six mistakes we see constantly, and the system that replaces them.
Mistake 1: One case, five problems
The classic first case from a frustrated founder: a paragraph covering the suppressed listing, the missing reimbursement, the flat-file error, and the Brand Registry glitch, ending with “please fix all of this.”
Here’s what happens mechanically. The associate — or the automation ahead of them — resolves the piece they can resolve, marks the case answered, and the other four problems evaporate. They weren’t declined. They were never registered as asks. You’ll re-raise them in a reply, the case will bounce between queues that each own one fragment, and three weeks later you’ll have a case with 14 replies and one-fifth of a resolution.
The rule: one case, one ask, stated in the first sentence. Not background first — the ask first. “Requesting reinstatement of ASIN B0XXXXXXX, suppressed on August 2 for a pesticide-claim flag that does not apply to this product. Details and evidence below.” Five problems means five cases, each cleanly scoped, each independently trackable. It feels slower. It is roughly three times faster.
Mistake 2: Filing without the evidence pack
Every round trip on a case costs 24–72 hours. A case opened without evidence guarantees at least two round trips before anyone can act: one for the associate to ask for what you should have attached, one for you to provide it — assuming they ask for the right thing, which they often don’t.
Write every case for a reader with four minutes and zero context. The evidence pack, attached at creation:
- The exact ASINs, SKUs, order IDs, or shipment IDs — never “several of our listings”
- Timestamps: when it broke, when you noticed, what changed around that date
- Screenshots of the error, the flag, the discrepancy — with dates visible
- Report exports that prove the claim (inventory ledger for missing units, the flat-file processing report for upload errors)
- What you already tried, so the first suggestion isn’t the thing you did before opening the case
There’s a 2026-specific reason this now matters more: the automated triage layer parses your case before routing it. Structured, specific, evidence-backed cases increasingly resolve without a human at all. Vague ones get a templated guess. You’re no longer just writing for a tired associate — you’re writing for a parser, and the parser rewards precision.
Mistake 3: Entering through the wrong door
Case categorization isn’t a formality — it’s routing. Pick the closest-looking menu path and you may land with a team that doesn’t have the tools to fix your actual problem. And here’s the behavior nobody warns you about: teams answer the question they can answer. A catalog issue misrouted to an FBA queue doesn’t get forwarded to catalog; it gets a polite, useless FBA-shaped answer, and the case gets closed as resolved.
If two rounds of replies make it obvious you’re talking to the wrong team, stop arguing with the queue. Re-file through the correct path with clean scoping — referencing the prior case ID for continuity — rather than trying to steer a misrouted case sideways. Arguing with the wrong team is the single largest source of case-age in the logs we audit.
Mistake 4: Accepting the first response — or letting the case die of natural causes
The first reply to a nontrivial case is very often a template: a restatement of general policy, a link to a help page, an answer to a question adjacent to yours. Sellers make one of two errors here. Some accept it and give up — logging the issue mentally as “Amazon won’t fix this.” Others get angry, fire back a paragraph of frustration with no new information, and get the same template again.
The move that wins cases is the precise reopen: “Your response addresses [X]. This case concerns [Y]. Per [specific policy / report / prior case #], the expected outcome is [Z]. Please escalate to the team that handles [Y].” No emotion, one new fact or reference if you have one, a clear restatement of the ask. In our experience the second or third round — conducted this way — is where the majority of nontrivial cases actually get resolved. The first response is a filter. Most sellers filter themselves out.
Two related disciplines: never let a case auto-close while the issue is live (an auto-closed case resets your progress and your paper trail), and never open a duplicate case because the first one stalled — duplicates fragment the history and can get both cases closed against each other. Reopen the original.
Mistake 5: Escalating emotionally instead of procedurally
There is an escalation ladder, and it works — in order:
What doesn’t work: all-caps, threats to call a lawyer, demands for supervisors in round one, and skipping straight to executive escalation for a routine issue. Your case history travels with the case. A clean procedural trail is itself evidence; an emotional one is noise the next reader has to dig through to find your actual problem.
Mistake 6: Running no case ledger at all
This is the structural one, and it’s the difference between operators and everyone else. Most brands have no record of their support history. Cases live in Seller Central’s interface, scattered and unsearchable in any useful way. The same issue gets rediscovered and refiled from scratch every quarter. The phrasing that worked to fix a variation split in March is in nobody’s notes when the same problem hits in September. And when the team changes — an employee leaves, an agency transitions out — the entire institutional memory of what’s broken, what’s pending, and what language works walks out the door with them.
The fix is a spreadsheet that takes five minutes a week: case ID, ASIN(s), issue class, date opened, current status, resolution, days-to-resolve — and a column for the exact language that worked. That last column compounds. Amazon’s queues respond to specific framings, and a brand that has learned its category’s magic words resolves repeat issues in one round instead of four.
The agency angle, since this is where we live: ask whoever manages your account to show you the case ledger. Not a summary — the ledger. A team that produces it in one click runs your account like an operation. A team that says “we handle cases as they come up” is telling you your support history is a group chat memory. We’ve audited accounts carrying 20+ resolvable issues — stranded inventory, dead variations, misapplied fee categories, stale suppressions — that no one could enumerate because no one had ever written them down. That list is usually worth real money, and it’s free to build.
The pre-Q4 sweep
Here’s the seasonal reason to fix this now, in August. Between now and October you have the last comfortable window to clear the backlog: queues are at normal speed, listings can be touched without peak-season risk, and fixes have time to settle before traffic arrives. The sweep we run on every account this month:
FAQ
How long should a case actually take?
Simple, well-evidenced cases: 24–72 hours. Nontrivial issues handled with clean escalation: one to three weeks. Anything past four weeks is almost always a case-hygiene failure — misrouting, fragmented asks, or a dead case nobody reopened — not Amazon refusing.
Does opening lots of cases hurt my account?
Volume itself, no. Duplicate cases on the same issue, yes — they fragment history and slow everything. One issue, one living case, reopened as needed.
When should I call instead of write?
Account health issues and anything with a deadline attached — call, with your case ID and evidence in front of you. For catalog and FBA issues, written cases with attachments generally beat calls because the paper trail persists and the evidence travels with the case.
Seller Assistant answered my question before I could file. Should I trust it?
Treat it as triage, not verdict. When it resolves the issue, done. When it gives you a policy summary that doesn’t match your situation, proceed to a structured human case — and note that its answer is now part of how you frame the ask (“automated guidance suggested X, which doesn’t apply because Y”).
Can my agency really move cases faster than I can?
A good one, yes — not through secret channels, but through pattern volume: they’ve seen your issue class fifty times and know the queue, the framing, and the evidence that resolves it in one round. That’s exactly what the ledger question in Mistake 6 tests.
—
If you’re looking for a team that manages every lever — creative, advertising, and operations — Velocity Sellers works with brands doing $100K+/month on Amazon. Contact us for a free account audit.
