04 Sept 2026 · The Bridge desk · 6 min read
What a night shift actually looks like
Ten hours, one account, ten decisions. The printout from 22:00 to 08:00, and what each entry cost to produce.
Read-only · seven days · no call
See seven days of decisions on your Amazon account
Connect one account read-only. The agents replay a week your account has already lived through, and you get back every decision they would have made, with the money attached to each one.
What a Decision Audit is
A Decision Audit is a replay. You connect one Amazon account read-only, the seven agents read the last seven days of history that account already holds, and you get back every decision they would have surfaced over that week — timestamped, costed, and with the reasoning chain attached to each one.
Nothing is changed while it runs. The audit is executed entirely in Watch mode, which is the first of the three permission modes, and Watch has no write path at all. There is no configuration step, no qualification call, and no trial that quietly starts billing on day fifteen.
The reason it works on history rather than on a live week is that the interesting failures are already in your data. A fee tier moved in August. A competitor undercut you at 02:40 on a Tuesday. Forty-one shipments reconciled short. None of those need to happen again for an agent to find them — they need somebody to assemble settlements, fee estimates, price history and the ad reports into the same view, which is the work nobody has time for.
How it runs, in four steps
One of these four needs you. The other three happen while you do something else.
- 01
Connect one account, read-only
You authorise Sellerbridge against a single seller account through Amazon's own Selling Partner and Advertising API consent screens. Read scopes only. No credentials are typed into this site, and no write scope is requested at any point during an audit.
About four minutes - 02
Seven days of history are pulled
Orders, settlements, fee estimates, FBA ledger movements, inbound shipments, listing status, buy box and price history, and the full advertising report set for the last seven days. Nothing older is requested and nothing is copied that the report set does not need.
Two to six hours, unattended - 03
The agents replay the week
Each agent runs its own beat across that history hour by hour, exactly as it would overnight on a live account. The Bridge then does the part that matters: collapsing findings that describe the same event into one decision instead of four separate alerts.
Runs straight through - 04
The report lands
One document, ordered by the money attached rather than by timestamp. Every decision carries what was observed, what was checked, what was ruled out, and which guardrail would have applied had the account been live.
Within one business day
What is read, and what is never touched
“Read-only” is only worth anything if it is itemised. Left column: the API surfaces an audit reads. Right column: what it does not do to them.
Advertising
Campaigns, ad groups, targets, search terms, placements, budgets and spend for the seven-day window.
No bid, budget, target or campaign state is changed.
Finance
Settlements, referral and FBA fee estimates, storage charges, reimbursements and returns.
Nothing is filed, claimed or disputed on your behalf.
Inventory
FBA ledger entries, inbound shipments, restock recommendations and days of cover.
No shipment is created and no purchase order is placed.
Catalogue
Listing status, suppressions, buy box share, price history and competitor price snapshots.
No listing content is edited and no price is repriced.
Account health
Policy and performance metrics as Amazon reports them.
No appeal or case is opened.
What lands at the end of it
One document. Ordered by the money attached rather than by timestamp, because the timestamp is not the part you act on.
Every decision, timestamped
The full seven-day printout, in the order the agents produced it, with the agent that raised each entry named.
The money attached to the ones you missed
Each decision carries a figure and how it was derived — spend against negative contribution, reimbursable loss, revenue exposed by a suppression, share lost to a price move.
The reasoning chain for each
Observed, checked, ruled out. The ruled-out line is the one worth reading: it is where you find out whether the agent considered the objection you were about to make.
The guardrail that would have applied
For every decision, what a live account in Watch, Recommend or Act would have done with it, so you can see what you would have been signing off before you sign anything off.
A read-only access log
Every API surface touched, when, and by which agent. Revoking access in Seller Central ends it immediately and permanently.
Before you connect anything
The four questions that come up before an audit, answered on the page rather than on a call.
Do I have to give write access?
No. An audit requests read scopes only, and the write scopes are a separate consent you would grant later, if ever. Watch mode has no write path, so there is nothing to switch off afterwards.
What if the seven days were quiet?
Then the report is short and says so. A week with three decisions in it is a legitimate result, and we would rather hand you that than pad the document. The commercial failure mode we care about is a decision that existed and was missed, not a thin week.
How many accounts can I connect?
One. An audit is deliberately a single account, because the point is to be checkable — you know that account well enough to tell us which of the decisions we surfaced were wrong. Agencies usually pick their worst-performing brand for exactly that reason.
What happens to the data afterwards?
It is used to produce your report and nothing else. No training, no pooling across sellers, no sharing with other tenants. Revoke access in Seller Central whenever you like; the report you were sent is yours to keep.
What an audit reads like
A worked night and the permission modes an audit runs under — both worth reading before you connect an account.
Connect one account, read-only. See seven days of what the agents would have caught.
That is the Decision Audit. No configuration, no write access, no call required to start.
Manifest — what lands in your inbox
01Every decision the seven agents would have surfaced, timestamped.
02The money attached to the ones you missed.
03The reasoning chain for each, so you can check our work.
04A read-only audit log of everything we touched.