How Brands Monitor MAP Violations on Amazon

"We monitor Amazon for MAP violations" is a sentence every brand with a pricing program says, and it describes a surprisingly wide range of activities — from an analyst opening product pages on a Friday afternoon to a scheduled system reading every offer on every listing several times a day. This article describes what the second kind involves, step by step, so that a brand evaluating its own program (or a tool) can see what each part is for.

Why Amazon is different

A brand's own website shows one price. A retailer's website shows one price. An Amazon product page shows offers — one listing, many sellers, each with its own price, condition, fulfillment method and shipping, one of them holding the Buy Box at any given moment. Fifteen sellers on a listing is unremarkable; fifty is not unusual for a popular product.

That structure decides what monitoring has to do. It is not enough to read "the price" on a page, because there is no such thing. Monitoring has to read every offer, record who made it, and measure each one separately.

Step one: read the offers, on a schedule

The first job is acquisition — getting the offers for a listing at a point in time. Two things matter about how it is done.

Through what channel. Amazon provides a sanctioned way for a selling partner to read listing offers programmatically: the Selling Partner API, authorized from the brand's own Seller Central account. Reading through it means the brand has granted the access, the access can be withdrawn from either side, and no marketplace access control is being worked around. The alternative — scraping the storefront while evading its anti-automation measures — produces observations a brand cannot comfortably present to anyone, because they are themselves evidence of a terms-of-service problem. MapProtector reads only through the authorization a brand grants; see Connect Amazon.

On what cadence. Prices on Amazon move. Repricing software adjusts a seller's offer several times a day in response to competitors; a violation that exists at 06:00 may be gone by 09:00 and back by noon. A scan once a week sees a sample; a scan every four hours sees the shape of the movement. Cadence is a cost decision and a plan decision, but it should be a deliberate one.

Step two: record what was seen — and how much

Each scan produces observations: one per offer, each recording the price and currency, the seller, the condition, the fulfillment channel, the Buy Box, and the instant. These are facts and they should never be edited afterwards.

The part most programs skip is recording coverage: how much of the listing the scan actually read. The offers endpoint may return a partial list, or the source may say it could not tell whether the list was complete. That matters for what a brand may conclude. A seller seen below the floor in a partial read was still seen; a seller absent from a partial read proves nothing — the seller may have been in the part that was not read.

A monitoring system that does not record coverage will, sooner or later, report a seller as compliant because a scan failed to see them, and then report them as "cured" for the same reason. Observations, coverage and what they permit are described in Observations.

Step three: identify the seller

Every offer names a seller, and the seller is the unit of enforcement. Two rules keep seller identification honest.

The identity is the marketplace's own seller identifier, not the storefront name. Storefronts rename themselves; the id persists. A program that keys on names will lose a seller's history the day they change their display name and will occasionally merge two different sellers that chose similar ones.

Identification is not classification. Knowing who a seller is says nothing about whether the brand has a relationship with them. A brand that keeps a list of its authorized resellers — its Authorized Network — can match each observed seller against it; a seller that matches nothing is unknown, which is a state, not a verdict. Deciding that an unknown seller is an unauthorized one is a person's call, made with a reason. Authorized vs Unauthorized Amazon Sellers explains why the distinction is load-bearing.

Step four: measure against the right policy version

An observed price is compared with the effective MAP at the instant of the observation. "Effective" carries a lot:

• the policy version in force on that date — a MAP raised in April does not retroactively turn March's compliant offers into violations, and a MAP lowered in April does not erase March's confirmed ones;
• any promotion the brand approved for a window covering the observation, which lowers the floor for that window;
• any exception — a distributor under negotiation, a product under legal review — that exempts the offer from judgment or bars enforcement while still recording the measurement;
• the treatment rules: whether shipping is included, how a visible coupon counts, whether a cart-only price is admissible, whether an out-of-stock offer is judged.

The outcome is one of a small set: compliant, below policy, indeterminate (the rules could not conclude — a couponed offer the policy flags for review, say), not applicable (an exception applied), not evaluated (no policy covered the product). Each one is distinct, and collapsing "indeterminate" or "not evaluated" into "compliant" inflates a compliance rate with offers nobody actually assessed.

Step five: decide what reaches a person

Not every below-policy observation deserves a person's attention. A program that puts all of them in front of a reviewer trains the reviewer to click through. Eligibility gates decide which ones become violation candidates: the seller was identified; the product is actively monitored rather than watch-only; no exception bars enforcement; the price was actually read; coverage was good enough to trust the presence; the observation is fresh.

Every gate should return a named reason, because "why is this seller not in my queue" is the most common question a monitoring program generates and the answer should be a fact rather than an investigation. In MapProtector those reasons are listed on Violation candidates.

Step six: a person decides

The candidate reaches a queue. A person — someone with the authority to say so on the brand's behalf — reads the observation, the seller's standing and the policy, and either confirms that the brand considers the price below its policy or dismisses the candidate with a reason from a closed list: it was not actually below policy; it was an authorized promotion; an exception covers it; the source data is not trusted; the observation is too old; the listing was matched to the wrong product; it duplicates another candidate.

The closed list is not bureaucracy. It is the only way a brand ever learns whether its monitoring is working: a month of dismissals for "wrong product mapping" is a catalog problem; a month of "authorized promotion" is a promotion that should have been configured before the sale; a month of "not actually below policy" is a treatment rule that needs revisiting.

Confirming opens a case — the persistent record of that seller on that product — and from there evidence, notices and verification follow. Those are the subjects of What Evidence Should You Capture for a MAP Violation? and How to Build a MAP Enforcement Workflow.

What monitoring is not

Monitoring is measurement. It does not decide that a seller breached an agreement, it does not decide that an unknown seller is unauthorized, and it does not send anything. A monitoring program that does any of those automatically has removed the person from exactly the decisions that will be quoted back to the brand by a seller with a lawyer.

The honest scope of a monitoring system is: read every offer through a channel the brand can defend; record each one with its coverage; identify the seller by the marketplace's id; measure against the policy version, promotion and exception in force; hold back what should not reach a person; and put the rest in front of one. The Amazon Seller Monitoring Guide covers the operational side — cadence, catalog hygiene, what to do in the first thirty days — in more depth.