Agencies21 August 2026 · 4 min read
Twenty clients, and nobody opening the morning sheet.
The agency version of this problem is not detection, it is distribution. Somebody already knows the listing moved. The question is whether the person who can act on it finds out the same day.
By SellerMate
Marrow Lane Commerce
Amazon agency
- Marketplace
- US, UK and DE
- Catalogue
- 21 client accounts
- Rules in place
- 3
- Manual checking
- 0 min/day90 min/day → 0 min/dayRoughly 32 working days a year of scrolling search results by hand.
- Weekend incident lag
- Same day~62 hours → ~9 hoursA Friday 19:00 price cut, from happening to reaching the account manager.
- Alerts read
- ~4/weekPer account manager, after routing. Few enough that each one still reads as news.
- The situation
- Every morning a junior checks each client’s main keywords by hand — open Amazon, search, scroll, note the position, compare against yesterday, paste into a shared sheet. About ninety minutes a day. The sheet is accurate and almost nobody reads it, so a Friday-evening price cut on a client’s hero ASIN typically surfaces on Monday afternoon.
- Rules in place
- Organic position falls below 10 → raise the bid, cappedCompetitor ASIN price drops more than 10%Star rating drops below 4.3
- How it plays out
- Each client’s trackers are addressed to that client’s account manager rather than to a shared list. The ninety minutes disappears, and — more usefully — the Friday-evening cut reaches the person who owns that account on Saturday morning instead of Monday. The junior’s time moves to the work the sheet was supposed to enable.
The sheet was never wrong. It was just addressed to everybody, which in practice means it was addressed to nobody. Routing it per account was a smaller change than we expected and it is the one that stuck.
The sheet was never the bottleneck
It is tempting to frame this as automation replacing manual work, and the ninety minutes is real. But the manual check was producing accurate information on time; it was producing it into a document nobody had a reason to open.
Every additional client made that worse. A sheet covering twenty accounts is mostly rows irrelevant to whoever is reading, and a reader who has learned that most rows are irrelevant stops scanning carefully.
Broadcasting to everyone is functionally the same as broadcasting to nobody. Relevance is what makes an alert get read.
One recipient list per client
The change is unglamorous: trackers are grouped by client, and each group is addressed to the one or two people who own that account. An account manager gets messages about their accounts and nothing else, so the base rate of “this concerns me” stays close to one.
WhatsApp does more work here than it does for a single seller. Agency account managers are frequently not at a desk, and a message reaching a phone on a Saturday morning is the difference between a same-day response and a Monday one.
Where the bid rule fits, and where it does not
The position rule is the one that can act: when a client’s organic position falls past ten, the bid on that keyword can go up, inside a ceiling set per client. It proposes rather than executes by default, so the account manager approves from the alert — which is the right posture on someone else’s ad spend.
The price and rating rules alert only. That is a product limit rather than a policy choice — auto-actions currently attach to position rules — but it is also the honest answer for an agency, because there is no single correct ad response to a client’s rating slipping.
The takeaway
Route by account, not to a shared channel. The reason monitoring stops working at agency scale is relevance, and relevance is a distribution setting rather than a detection problem.
Last updated 23 August 2026
Set the rule while you are thinking about it.
Check any listing free. No Amazon login, no card.
Check a listing