In This Article
Most Amazon sellers can tell you their ACoS to two decimal places but have no idea how much they lost to A-to-Z guarantee claims last quarter. The money leaves quietly. A customer claims an order never arrived, Amazon refunds them from the guarantee fund, and the seller absorbs the cost without ever checking whether the claim could have been challenged.
This case study covers a mid-sized FBA seller who came to us losing £4,000 to £5,000 a month to exactly this. Over six months, the automated monitoring and dispute system we built recovered roughly £40,000 in revenue that would otherwise have been written off. Here is how the leak developed, what we built to plug it, and what happened next.
Load: The Silent Revenue Leak
Running a successful FBA business means juggling inventory, pricing, customer service and a hundred other moving parts. Somewhere near the bottom of that list sits the job of disputing A-to-Z guarantee claims through SafeT. For most sellers it never gets reached at all.
The mechanics matter here. When a customer files an A-to-Z claim against an order, you have a limited window to respond, typically three to seven days depending on the claim type. Your response needs evidence: tracking information, signature confirmation, carrier delivery proof, or communication logs showing the issue was already resolved. Miss the window, or file something vague, and the refund stands.
Our client had no systematic way of monitoring incoming claims. Some were missed entirely because the notifications sat in a shared inbox alongside order confirmations and review alerts. Others were spotted with a day to spare, which meant a rushed response built from whatever evidence was to hand. The rest were simply ignored, because the seller believed SafeT was an esoteric process reserved for complicated edge cases.
None of this was customer fraud. It was a breakdown in the seller's ability to defend their own transactions. When we audited their claim history, the vast majority were easily defensible. Item-not-received claims where tracking showed a signed delivery. Return-not-received claims where nothing had ever been scanned back into the warehouse. Legitimate sales, refunded through Amazon's guarantee, with the cost landing on the seller month after month.
Fulcrum: Building the Automated Defence System
The core issue was never technical complexity. It was visibility and process discipline. The seller needed to know about every claim the moment it arrived, have the evidence assembled without hunting through five different systems, and submit responses that actually addressed what each claim alleged.
We started with monitoring. A daily automated feed pulls every open A-to-Z claim from Seller Central and places it on a prioritised dashboard showing the claim type, customer name, order number, claim date and, most importantly, the response deadline. Nothing hides in a notification email any more. Anything with a deadline inside 48 hours moves to the top and triggers a direct alert.
Evidence gathering was the part that changed the daily reality most. When a claim lands, the automation pulls supporting documentation from every relevant source: FBA shipment records including weight, dimensions and condition photos, carrier tracking data and delivery confirmation, customer service logs showing whether the buyer had messaged and what was said, refund history and transaction details, and return label information where a return was involved. All of it is compiled into a single document, organised in the order Amazon's review team will want to see it. What used to mean half an hour of tab-hopping per claim now happens before anyone opens the case.
The final piece was structured submission. Amazon weighs evidence differently depending on the claim type, so we built guided templates for each one. An item-not-received claim leads with delivery proof. A significantly-not-as-described claim leads with listing details and condition records. The system populates the relevant sections automatically and flags any gaps that need a human decision before anything gets filed.
That last point matters. The seller still reviews and approves every submission. The difference is that they are approving a complete, professional package rather than scrambling to build a case from scratch at 11:58pm on day three.
Lift: Results and Recovery
In the first month the system caught 34 missed claims and recovered £8,200. The dispute success rate rose from roughly 45% to 78%, and by month six cumulative recovery stood at approximately £40,000.
The first month set the tone. The system caught and processed 34 open claims the seller had missed entirely, recovering £8,200 from those alone. That paid for the build several times over before the new process had even settled in.
Over the following three months the system was handling 80 to 100 new claims monthly. The dispute success rate climbed from roughly 45%, which is about the historical average for sellers without a system, to 78%. Nothing aggressive was going on. The improvement came from presenting complete evidence instead of vague explanations, every time, on every claim.
By month six, cumulative recovery stood at approximately £40,000.
The money was the headline, but three quieter benefits followed. Deadlines stopped being a source of stress, because nothing depended on someone spotting an email in time. Account health improved, since Amazon tracks claim outcomes in its seller metrics and a higher success rate feeds directly into them. And the dashboard surfaced patterns nobody had seen before: certain products carried noticeably higher claim rates, and certain carrier routes generated far more item-not-received claims than others. The seller used that data to adjust packaging on the problem products and move volume away from the weakest carrier route.
What SafeT Actually Is
For sellers meeting the term for the first time, SafeT is Amazon's formal dispute mechanism within the A-to-Z Guarantee system. Customers use A-to-Z to claim refunds. Sellers use SafeT to respond and challenge those claims with evidence.
Item-not-received claims are the most common and usually the most defensible, because tracking and delivery confirmation carry real weight with Amazon's reviewers. Significantly-not-as-described claims turn on condition, specification and colour, so listing accuracy and shipment records decide them. Unauthorised transaction claims are rare unless account security has been compromised. And return-not-received claims apply when a customer says they sent something back that never reached your warehouse.
Two things decide most cases. You must respond inside the deadline window, and your evidence must address the specific claim type. A generic "we shipped it" response convinces nobody. Amazon's team reviews each case on the evidence in front of them, and thorough, professional submissions win most of the time. The bigger problem is that many sellers never respond at all, because they assume Amazon's decision is final. It is not.
Why Automation Made the Difference
Manual SafeT management fails in three predictable ways. Claims get lost in notification noise. Responses get rushed and arrive incomplete. And attention is inconsistent, so some claims get a proper defence while others get nothing. Automation removes all three failure modes at once, because the process runs whether or not anyone remembered to check.
It is worth being honest about where AI fits in this system: it does not. Claim monitoring, evidence gathering and templated submission are rules, schedules and structured data, and that is exactly the right tool for the job. Adding a model would have made it more expensive without making it any better.
When our client saw the first quarter's numbers, their reaction was "why isn't every seller doing this?" The answer is that it requires systems and process thinking. It is not glamorous and it does not increase sales. What it does is protect revenue you have already earned, which for an FBA business handling hundreds of orders a month is often the cheapest margin available.
If you suspect you have the same leak, that is what Fulcrum Three does. We design, build and run this kind of system as part of a managed operations engagement, and A-to-Z recovery is one of the fastest places to prove the value.
Ready to recover lost revenue from A-to-Z claims?
Book a Free Operations Audit →