Account health: building a system, not chasing fires

Most sellers only look at their account health dashboard after something's already broken, a warning notice, a suspended listing, an email that starts with "Your account may be at risk." By that point, whatever caused it has usually been building for weeks, and reacting to a single flag is a different, harder problem than catching the trend before it became a flag at all.

A real monitoring system checks the same handful of metrics on a fixed schedule, before there's any reason to, so problems get caught while they're still small and easy to fix.

The metrics that actually matter, and why

Order Defect Rate combines negative feedback, A-to-z claims, and chargebacks into one percentage, and Amazon's threshold for this one is unforgiving: even a small number of defects on a low-volume account can push the percentage above the acceptable range, since it's a rate against total orders, not an absolute count. A new or smaller seller can trip this threshold with just a handful of problem orders in a short window.

Late Shipment Rate matters because Amazon treats delivery reliability as a core trust signal, and this one is often affected by external factors, carrier delays, holiday volume, supplier issues, that aren't fully within a seller's control, which makes it worth tracking even when nothing feels obviously wrong on your end.

Valid Tracking Rate and Cancellation Rate get less attention than ODR but contribute to the same overall account health picture, and a slow drift in either one over several weeks is a pattern worth catching before it compounds with something else.

Policy violation notices, even minor ones that don't immediately threaten the account, are worth logging over time rather than dismissed individually, since a pattern of small violations across different areas reads very differently to Amazon than one isolated incident, even if no single one looked serious when it happened.

Why weekly beats monthly

Metrics that are fine on a monthly average can hide a bad two-week stretch that's actually driving toward a threshold, especially since some Amazon metrics are calculated on a rolling window rather than reset monthly. A defect rate creeping up during weeks three and four gets diluted by two good weeks earlier in the same month if you're only checking once a month, and the trend isn't visible until it's already close to the threshold.

Checking weekly catches the same information while there's still time to identify what's driving the trend and adjust before it becomes a genuine risk to the account, rather than after Amazon has already flagged it.

What weekly monitoring actually looks like in practice

Pull the account health dashboard on the same day each week and record the core metrics somewhere you can see the trend over time, not just the current snapshot, since the dashboard itself shows current standing more clearly than historical movement. A simple spreadsheet tracking each metric week over week makes a slow drift visible in a way that checking the dashboard alone often doesn't.

Cross-reference against what actually happened that week: a returned shipment, a supplier delay, a specific customer complaint. Connecting a metric change to its actual cause, while the cause is still fresh, makes it much easier to fix the underlying issue rather than just watching a number move without understanding why.

Flag anything trending in the wrong direction for two consecutive weeks, even if it's still well within Amazon's acceptable range. Two weeks of the same direction is usually a real pattern rather than normal noise, and catching it there means addressing a small drift instead of an already-serious problem.

The role of a named owner

Account health monitoring that's "everyone's job" tends to become no one's job, especially once a business has more urgent things demanding attention in a given week. Naming a specific person responsible for the weekly check, even in a small team where that person wears several hats, makes the difference between monitoring that actually happens consistently and monitoring that quietly stops the first time something more urgent comes up.

For solo sellers, this just means putting the check on a recurring calendar reminder rather than relying on remembering to look, since the moment it's easy to skip is exactly the moment something's more likely to be building underneath the surface.

What to do when a metric starts trending wrong

Identify the specific orders or listings driving the change, not just the aggregate percentage. If ODR is climbing, which specific orders generated the negative feedback or claims, and is there a common thread, a specific product, a specific shipping method, a specific time period where something changed upstream.

Address the specific cause rather than a general effort to "be more careful." If a particular product's return rate is driving the trend, that points toward a listing accuracy problem, a product quality issue, or expectations set incorrectly in the listing itself, three different problems with three different fixes.

Document what you found and what you changed, even informally. If Amazon ever does flag the account later for a related issue, having a record showing you'd already identified and addressed the pattern proactively is meaningfully different from having no documentation of ever noticing it.

Why this is cheaper than it looks

Weekly monitoring takes a fraction of the time that reacting to an actual suspension does. A suspension typically means the affected listings stop generating sales entirely during the appeal process, plus the time cost of building a Plan of Action, gathering evidence, and waiting for review, often measured in days to weeks of lost revenue. Fifteen minutes a week checking four metrics is a small cost against that, and it's a cost you control the timing of, rather than one Amazon imposes on you without warning.