Data health and sync

How syncing works

The extension syncs in the background of your browser, in short time-boxed slices, and resumes where it left off. A closed laptop or a killed service worker costs nothing but time.

Three lanes run together: listings, sales (newest days first) and advertising (the whole campaign → ad group → target hierarchy). Slower work — deep sales history, ads detail, deleted listings — runs behind them as reconciliation, and the screens name it separately so you can tell a sync that is finishing from one that is still arriving.

After the first sync, updates are incremental:

There is no server doing this for you. Syncs run in your browser, so a browser that is closed is a browser that is not syncing. It catches up the next time Chrome is open.

The two screens

They are different screens with similar names, and it is worth knowing which one answers which question.

Data status is the line at the foot of the sidebar. It says what is happening right now: which lanes are working, which are paused, and which reconciliation passes are still completing. It reads Fetching data while anything is active and Data status when nothing is.

Data health is a tab. It says whether what has already been fetched adds up: Summary rebuilds, Rebuild backlog, and Diagnostics. It also has the download button you would send us if something looks wrong, because the file names exactly what the extension knows and does not.

Three rules govern what you see there, and all three exist so that a gap is never mistaken for a zero:

A failed day retries itself. A day that failed — a rate limit, a dropped connection — stays in the missing set and comes back round on a later pass. Failed days are never silently skipped.

Unchanged days are skipped. Every fetched day carries a fingerprint, so re-checking history costs almost nothing and the extension can afford to keep checking.

Rebuilds land whole. A read that happens in the middle of a rebuild can never show numbers that are half old and half new.

Why numbers move, and when they stop

A day is complete once it has ended in Los Angeles — the timezone Amazon's reporting uses for every marketplace — and been fetched afterwards. Today's numbers move; yesterday's usually do not.

"Usually" is doing real work in that sentence. A cancellation or a return is booked on the day it is observed, not on the day the original order was placed, so a finished day can still change when something that happened later lands on it.

Comparisons only ever use equal counts of complete days, which is why today never gets one.

The advertising window

Amazon serves roughly the last 90 days of daily advertising data. The window rolls: it always ends today, and it starts at a different point in each marketplace, because each marketplace's advertising was connected at its own moment. Older history arrives at month grain through a separate one-time backfill. Search terms reach back a shorter 65 days, and have no month-grain history behind them at all.

Two years of sales beside three months of ad spend is therefore the data Amazon gives, not a failed fetch.

Past the window, a spend of zero means "not known" — not "no ads ran". The dashboard is careful never to present the first as the second, and the two are shown differently wherever the distinction matters.

One machine, one database

The database lives in this browser profile. A second computer starts its own sync from zero, and clearing the browser's storage clears it — the history re-syncs from Amazon, slowly. There is no cloud copy of any of it.

Two things worth knowing before you rely on that:

• An extension update that changes the shape of the database rebuilds it from scratch. Nothing is lost that Amazon still serves, but the long first sync happens again.
• Because the sync runs in your browser, the machine you keep open is the machine with the most complete history.

If that trade is new to you, what MerchPPC is sets it out in both directions.