How to Track Repeat MAP Violations

The seller who matters most in a MAP program is rarely the one with the deepest single discount. It is the one who drops below the floor, raises the price when contacted, and drops again three weeks later. That pattern is what distinguishes a mistake from a strategy, and it is exactly the pattern most tracking methods lose — because they close the finding when the price goes back up, and the next drop looks like a first offense.

This article describes a way of tracking repeat violations that keeps the history intact: cases, breach periods, a cure rule that refuses to be fooled, and a recurrence count that means what it says.

The unit is the relationship, not the incident

Most programs track incidents: a violation is opened, handled and closed, and the next one is a new row. That loses the thread. Ask "has this seller done this before?" and the answer requires someone to search old rows by a name that may have changed.

The alternative is to track the relationship: one record — a case — for one seller on one product under a policy, which persists whether or not anything is currently wrong. Inside it, each continuous stretch of non-compliance is a breach period with its own start, end and outcome. A case does not close when the price goes back up; the period does, and the case rests at monitoring until the next period opens on it.

With that structure, "has this seller done this before?" is a count of periods on the case. Period 3 is the third breach.

What opens a period, and what extends it

A period opens when a person confirms a below-policy finding for a seller and product that has no open period. Nothing else opens one — not a scan, not a worker, not a schedule. If the case already has an open period, the confirmation attaches to it and the period continues.

Every later observation of that seller on that product attaches to the open period: the below-floor ones, which extend it, and the compliant ones, which count toward its cure. Both are kept. A period that recorded only the observations that kept it open would never be able to close.

The period's running figures — latest price, latest floor, latest gap, observation count — move only forward. A replayed older observation cannot overwrite a newer price.

What closes a period — and what must not

Here is where most tracking goes wrong. The tempting rule is "close it when the seller is no longer below the floor". The problem is what "no longer below" gets measured from.

• The seller disappeared from the listing. Not a cure: they may be in the part the scan did not read, or out of stock, or back tomorrow.
• The price went hidden — "see price in cart". Not a cure: a hidden price is not a raised one.
• The scan failed. Not a cure: nothing was observed.
• One compliant reading. Weak: a repricer flicker or a Buy Box rotation can produce a single at-floor reading between two below-floor ones.
• The seller wrote back saying they have fixed it. Not a cure: a claim is a claim.

Each of these, treated as a cure, closes the period wrongly. And a wrongly closed period does not just lose a day — it makes the seller's next below-floor offer look like a first offense. A false cure launders a seller's history.

The rule that holds is stated in the positive: a period closes as cured when fresh, sufficiently covered observations of the same seller on the same product show the price at or above the floor — enough of them, far enough apart. Three settings control how much evidence; nothing controls what kind.

A below-floor observation resets the run. An observation that is neither compliant nor below floor — hidden, failed, partial — neither resets nor extends it; it is silence, and silence proves nothing either way. MapProtector's Case lifecycle page lists every reason a period will not close, in the words shown on the case.

Cured and closed are different words

A period can also end because a person ended it, or because the product was archived. Those closures must be recorded as closed, not cured. Cured is a fact about the seller — the price was observed back at the floor. Closed is a fact about the brand — someone set it aside. Merging them would let an administrative tidy-up read, in the seller's history, as the seller correcting themselves.

Counting recurrence honestly

Once periods are immutable and closures are honest, recurrence is derived from the chain rather than stored: how many periods the case has, how many beyond the first, the hours between the previous closure and the latest opening, and whether one is running now.

Across a seller's cases, the count sums periods beyond the first, per case. Consider two sellers, each with three periods in total:

• Seller A: three cases, one period each. Three separate first instances on three products. Recurrence: zero.
• Seller B: one case, three periods. One product, corrected and relapsed twice. Recurrence: two.

A naive count calls them the same. Only Seller B is a repeat offender, and a program that cannot tell them apart will escalate against the wrong one.

What the record looks like

On the seller's page: products observed, active and historical periods, confirmed and dismissed findings, products currently below policy, and the recurrence count. On the case: the current period, why it has not closed, every notice for the period, responses, the timeline, and the previous periods with their outcomes. On the queue: a sort by repeat seller, which orders one measured quantity by itself rather than inventing a composite score.

What a relapse should trigger

Nothing automatic. A new period on a case with history is a fact the program surfaces; whether it warrants a stronger notice, a call to the distributor, or a review of the seller's standing is a person's decision. The escalation sequence a brand runs is chosen when someone decides to act on the case again — and a program that escalated on its own because a seller relapsed would eventually escalate against a partner whose repricer misfired twice. How to Build a MAP Enforcement Workflow covers where the person sits in that sequence.

The short version: track the relationship, not the incident; close a period only on observed cure; keep closed periods immutable; count recurrence per case. A history kept that way is one you can act on.