How the store reporting pulse works
Every Monday, a busy store owner gets one branded page that says what happened last week, what moved, and what deserves attention, the kind of read you'd want from a sharp analyst, in about five minutes. No dashboards to dig through, no numbers to assemble.
A look under the hood: where the data comes from, what the weekly report covers, and the strict rule that every number traces to the source.
The short version
It pulls your store's analytics, sales, orders, average order value, traffic, where it came from, top products, discounts, and writes a weekly performance report a busy owner reads in five minutes. A small daily job keeps a fresh sales snapshot; a bigger weekly pull gathers everything; then an analyst agent reads it all and writes the report into a fixed, branded page, same structure every week, with last week compared to the week before and to the same week a year ago. It's a strategic read, not an operations to-do list, and it never changes a thing in your store.
That's the whole job. The rest of this page is where the numbers come from, and the rule that keeps the report honest.
A week of store data, read down to five minutes. The pull is exact and read-only; the green step is the analyst's read; the report is the same branded page every Monday.
Everything the store can tell you, read-only
It reads your store's own analytics directly, and only reads. Revenue, orders, average order value, profit, discounts, returns, new versus returning customers, sessions and conversion rate, traffic sources, top products, sales by country. No customer personal data, ever, just the aggregates.
Two scheduled jobs do the gathering, and both are plain, deterministic code, no AI, no tokens. A daily one keeps a current sales snapshot; a weekly one, run over the weekend, pulls everything the report needs into organised files with a note on how fresh each piece is. Because it only reads, it can never alter an order, a price, or a product, it's a window onto your store, not a hand in it. And it never touches customer personal information; the whole report is built from aggregate numbers.
A read, not just a dashboard
The weekly report is where the analyst earns its keep: it doesn't just show the numbers, it reads them. What grew, what dropped, how this week compares to last week and to a year ago, and which two or three things actually deserve the owner's attention.
It lands Monday morning as a polished, password-protected page at a permanent link, same address every week, fresh content, plus a three-or-four-line summary in chat that leads with the revenue number. It covers the last full week, compares it to the previous week and to the same calendar window last year, and if the data is stale or anything went wrong gathering it, it says so right at the top rather than quietly reporting on a hole. The structure is identical every week, so the owner always knows where to look.
Every number traces to the source
The analyst writes the words and picks what matters, but it is not allowed to invent a single figure. Every number in the report comes from the gathered data, and if a value is missing, the report flags it rather than letting the analyst fill the gap with a guess.
This is the rule that makes a written report trustworthy: the judgment is the model's, the facts are the store's. The figures are computed from the read-only data and slotted into a strict template; a missing value fails the build instead of being papered over. So you get the readability of a human-written brief with the reliability of a spreadsheet, the analysis is fresh each week, but no number in it is ever estimated or remembered.
The owner gets the readability of a human brief and the reliability of a ledger, because the words are written fresh and the numbers never are.