Guide
Amazon Seller Monitoring Guide
The operational side of monitoring Amazon for a MAP program — how offers are read, what to record, how to set cadence, what coverage means, how to keep the catalog clean, and how to read the first month's results.
This guide covers monitoring as an operation: what has to be true for a scan to mean something, how to set it up, and how to read what it produces. It is written for the person who will run the program day to day. The reasoning behind the design choices is in How Brands Monitor MAP Violations on Amazon.
1. Acquisition: how offers are read
An Amazon product page is a set of offers from many sellers, not a price. Monitoring reads all of them for a listing at a point in time.
Channel. Read through the Selling Partner API, authorized from the brand's own Seller Central account. The brand grants the access; the brand can withdraw it; nothing works around the marketplace's controls. Observations obtained by scraping around anti-automation measures are observations you may not want to present to anyone. In MapProtector the authorization is a one-time consent from Seller Central; see Connect Amazon.
What is read. Listing offers and the identity of the seller behind each — not orders, customers, inventory or financial data. A monitoring authorization should ask for no more than that, and a program should be able to say so to its own security team.
Marketplaces. The authorization covers the marketplaces the brand approves. A listing on a marketplace outside the authorization will not be read; check the connection's marketplaces against the catalog's.
2. Observations: what to record
Each scan produces one observation per offer. Record, and never edit afterwards:
| Field | Note |
|---|---|
| Seller id and storefront name | The id is the identity; the name is how it looked that day. |
| Price and currency | Exactly as reported. A hidden or cart-only price is recorded as price not shown, never as zero. |
| Condition, fulfillment, Buy Box | New/used/refurbished; who ships; whether the offer held the Buy Box at that instant. |
| Instant | UTC. |
| Coverage | Sufficient, partial, insufficient or unknown — see §4. |
| Evaluation | The outcome and the policy version used — see §6. |
The rule that makes observations trustworthy is that they are facts written once. A program that "cleans up" old observations has no history.
3. Cadence: how often to look
Cadence is a cost and a plan decision, and it should be a deliberate one. Three considerations:
- Repricers. Automated repricing moves a seller's offer several times a day. Once a day sees a sample; every four hours sees the shape.
- Freshness. A finding becomes a candidate only while the observation is fresh, and a notice can cite only a fresh observation. Cadence keeps findings inside those windows.
- Cure. A breach period closes only on a run of compliant observations far enough apart; at twice a day a corrected price is recorded as cured within roughly a day, at once a week within a fortnight.
MapProtector's plans run twice a day (Starter), every six hours (Growth) and every four (Professional); Enterprise is agreed per contract. Schedules are weekly slots in your timezone, so a weekend can differ. For a listing you are watching closely today, an on-demand scan queues immediately without changing the schedule. See Monitoring cadence.
4. Coverage: what an absence means
A scan reports how much of the listing it could see. This is the most important field most programs do not record.
| Coverage | A seller seen below the floor | A seller not seen |
|---|---|---|
| Sufficient | Was seen below the floor. | Was, as far as the source can tell, not offering. |
| Partial | Was seen below the floor. | Proves nothing — may be in the part not read. |
| Insufficient / unknown | Not trustworthy enough to act on. | Proves nothing. |
Two consequences follow. Partial coverage does not hold back a below-floor observation from review — a presence was seen. Partial coverage does stop a breach period from curing — a cure needs the whole picture. A program that gets this asymmetry backwards either misses real findings or manufactures false cures. Observations sets out the rule.
5. Catalog hygiene
Most "false positives" in the first month are catalog problems. Before blaming the monitoring:
- MAP in the MAP column. Not MSRP, not the brand's own selling price. MAP vs MSRP explains the mistake.
- The right ASIN. A product mapped to the wrong listing produces findings about a different product. The dismissal reason matched to the wrong product exists for this; a month of it is a catalog problem.
- Condition. A policy for new condition should not judge used or refurbished offers. Set the condition on the product and the policy.
- Marketplace. A product listed in a marketplace the authorization does not cover will read never scanned.
- Lifecycle. Active products are judged and reach the queue. Monitor only products are judged but never reach it. Discontinued products still qualify — a liquidator below the floor on a discontinued line is a pattern worth seeing. Archived products are not scanned.
- Import by file, validate first. For a catalog of any size, import a CSV, read the row report, correct, then apply. Valid rows load even when others fail, and a file already imported is recognized.
6. Evaluation outcomes
Every observation carries one of six outcomes, and a program should report them separately rather than folding the non-answers into "compliant".
| Outcome | Meaning |
|---|---|
| Compliant | At or above the effective floor, or inside the tolerance. |
| Below policy | Under the effective floor after tolerance and treatment rules. Eligible ones become candidates. |
| Indeterminate | The rules could not conclude — shipping required but not shown, a couponed offer flagged for review, a cart-only price under the default rule. |
| Not applicable | An exception exempted the offer. Not judged — not the same as compliant. |
| Not evaluated | No policy covered the product at that instant. |
| Price not shown | The marketplace did not display a readable price. |
A compliance rate that counts not applicable or not evaluated as compliant is flattered. Ask any tool what it does with them.
7. Seller identity
The identity is the marketplace's seller id. Storefront names are aliases: a rename keeps the history and the old name is kept for recognition. Nothing merges two ids into one seller on the strength of a name, an address or a price, and nothing should.
Every newly seen seller is matched against the Authorized Network on first sight. A match is authorized; a non-match is unknown — a state, not a verdict. The Authorized Reseller Management Guide covers the list; Seller discovery covers what happens on first sight.
8. Reading the first month
Week one. Expect never scanned on products whose marketplace is outside the authorization, not evaluated on products no policy covers yet, and a long unknown-seller list. None of these is a finding about compliance; all of them are configuration.
Week two. The queue fills. Before working it, check the dismissal reasons you find yourself using. Wrong product mapping is the catalog; authorized promotion is a sale that should have been configured; not actually below policy is usually a treatment rule (shipping, coupons) that the written policy and the configured policy disagree about.
Week three. Work the unknown sellers. Most will be partners or partners' customers. Add them to the network; the queue's authorization column starts to mean something.
Week four. Read the overview honestly. The compliance chart should show gaps on days with no successful scan — a hatched gap, not a zero. The unknown count should be falling. Below-floor findings should be visibly partner or stranger, and handled differently.
9. What monitoring does not do
It measures. It does not decide a seller is unauthorized, it does not decide a breach of agreement, and it does not send anything. A monitoring tool that does any of these automatically has taken a decision the brand will have to defend without having made it. The rest of the workflow — review, evidence, enforcement, verification — is covered in the MAP Enforcement Workflow Guide.