How to automate a weekly Shopify sales report

The weekly report is slow because the data lives in several places. Orders and sales are in Shopify, stock may be in an inventory tool or a spreadsheet, ad spend sits in each ad platform, and refunds show up days after the order. Someone exports each one, pastes them together, fixes the columns that do not line up, and builds the same charts again.

By the time it is read it is already out of date, and because it is rebuilt by hand, the definitions drift. One week 'sales' includes shipping, the next it does not. Abid Co's owner put it plainly: every week someone lost a day pulling sales into a spreadsheet.

A useful weekly report answers a few questions quickly. How did this week compare with last week and with the same week last year? Which products sold best, and which barely moved? How many weeks of stock are left on the best sellers at the current rate of sale? What came back as returns, and why? Anything that does not change a decision about stock, pricing, or marketing can go.

The comparison against the same week last year matters more than it looks, because it strips out seasonality. A strong week in December tells you little on its own. A strong week against last December tells you something.

The lightest option is Shopify's own analytics plus a scheduled export or a Shopify Flow automation that sends key numbers each week. It costs almost nothing and suits a single store with simple needs, but it struggles once stock, ads, or a second sales channel need to sit in the same report.

The middle option connects Shopify to a spreadsheet or dashboard tool, such as Google Sheets or Looker Studio, through a connector that refreshes on a schedule. This handles most small stores well and stays readable by anyone on the team.

The fullest option is a small data pipeline. Shopify, inventory, and ad data land in one database on a schedule, get cleaned into consistent tables, and feed a dashboard and a written summary. It takes more to set up, and it is the one that keeps working when the business adds a warehouse, a marketplace, or a second store.

Automating a report with inconsistent definitions just produces wrong numbers faster. Decide once what 'sales' means: before or after discounts, with or without tax and shipping, counted on the order date or the payment date. Decide how refunds are handled and which week they land in. Write those rules down and build them into the pipeline, so the report means the same thing every week.

This is the unglamorous part, and it is the same work that sits under every data project. We wrote about why it comes first in most of the AI work is the pipeline no one wanted to build.

Once the numbers are reliable, a language model can turn them into a short written summary: what went up, what went down, which products to reorder, and which to consider dropping. That is the part people actually read, and it saves another chunk of the weekly effort.

Keep the arithmetic out of the model. Compute every figure in the pipeline or a query, and give the model only the finished numbers to describe. Models are good at explaining a table and unreliable at adding one up, and a sales report is the wrong place to find out which one you got.

For Abid Co we built a reporting layer that pulls sales, stock, and orders into one live dashboard and writes the weekly summary automatically, with best sellers and slow movers flagged each week. The six hours a week of manual reporting went to zero, and in the owner's words, they now look at the numbers instead of building them. It is the most common first project in our data engineering and analytics work, and one of several store jobs covered in automating an ecommerce store.