Refresh your store's homepage every Monday based on what sold

Who it is for

Online store owners, on Shopify, WooCommerce or similar, who know the homepage matters but change it "when there is time", which means almost never.

The problem

An outdated homepage shows what already sold out and hides what is selling today. Updating it means looking at sales, choosing products, writing copy and editing the store: manual work that gets postponed every week.

Ingredients

• Sales export for the last four weeks (product, units, amount)
• Your current stock (so you do not feature what is sold out)
• Your commercial calendar for the month: dates, promotions, launches
• An account on claude.ai or ChatGPT; optionally, n8n to automate the Monday delivery

Step by step

1. 1

   Gather the three files

   Four weeks of sales, current stock and the month's calendar. It does not matter whether they are CSVs from the platform or a hand-made spreadsheet; what matters is that the columns have the same names every week.

2. 2

   Write down your store's rules

   What never goes on the homepage (low-margin products, what arrives next week), how many products fit, what tone the brand uses. These rules are the part of the prompt that improves the result the most.

3. 3

   Ask for the Monday proposal

   Attach the files and run the prompt. Check that it does not propose anything sold out and that the reason for each choice comes from data ("up 60% in two weeks"), not from taste.

4. 4

   Approve, publish and note what happened

   Publish the changes in your store. The following week, add one line to the prompt: "last week we featured X and Y happened". The model learns from your own history without you programming anything.

5. 5

   Automate once the routine works

   Once you have done it by hand for a month and trust the result, an n8n workflow can export the sales, call the API with this prompt and email you the proposal on Monday at 8. Not before: automating something that does not work yet only makes it fail faster.

Replace what is in brackets with your company details, or use the guided mode above: it asks you and fills it in for you.

# Role
You are the merchandising manager at [STORE NAME], an online store that sells [WHAT IT SELLS] in Chile. You decide with data, not with taste, and you write in the brand's tone: [BRAND TONE, FOR EXAMPLE "warm and direct"].

# Context
Attached are three files: sales for the last four weeks (product, units, amount), current stock and the commercial calendar for the month. Store rules: the homepage fits [NUMBER OF PRODUCTS ON THE HOMEPAGE] products; nothing with stock below five units ever goes on the homepage, nor [PRODUCTS THAT NEVER GO ON THE HOMEPAGE, OR "NONE"]. Last week we featured: [WHAT WE FEATURED LAST WEEK AND WHAT HAPPENED, OR "THIS IS THE FIRST TIME"].

# Task
Propose this week's homepage: which products to feature and in what order, a promotion consistent with the calendar, and the copy for the main banner (title of at most 6 words and subtitle of at most 20). Each choice must cite the data point that justifies it.

# Output format
1) Table: position · product · data point that justifies it · stock. 2) Proposed promotion and why this week. 3) Banner copy. 4) Two things you would NOT do this week and why. If any file is missing or incomplete, say so before proposing.

# Examples
Input: "Basic black t-shirt: 12 units week 1, 19 week 2, 31 week 3, 44 week 4; stock 120".
Expected output: "Position 1 · Basic black t-shirt · sales almost quadrupled in four weeks (12 → 44) · stock 120, enough".

What usually goes wrong

• ✕Giving it only the sales without the stock. It proposes featuring the star product that sold out yesterday.
• ✕Changing the file format every week. The prompt stops understanding the columns and the proposals lose quality without you noticing why.
• ✕Automating before the manual routine works. If the Monday proposal still needs fixing by hand, the automated workflow just sends you bad proposals earlier.