Forecast next month's sales from your spreadsheet, without an analyst

Who it is for

Owners and people in charge of purchasing or shifts in retail, food, services: those who decide how much to buy or how many people to schedule next month by looking at last month.

The problem

Buying and staffing are decided by eye. In high months there is not enough stock or staff; in low months there is too much. The sales history sits in a spreadsheet and nobody looks beyond last month.

Ingredients

• Monthly sales history of at least two years, one row per month, one column per product or branch
• Google Sheets (the README mentions TimesFM being available there) or, for the code path, Python
• Fifteen minutes with whoever decides purchases to agree on what will be done with the forecast

Step by step

1. 1

   Arrange the history as a time series

   One row per month, no gaps: if you sold nothing one month, put zero, not blank. One column per product or branch. Dates in the first column.

2. 2

   Generate the forecast

   In Google Sheets, apply the TimesFM-based forecasting function to each sales column for the next one to three months. If you prefer code, the README includes the installation and a minimal Python example.

3. 3

   Compare against last year

   Put the forecast next to the same month of the previous year and the average of the last three months. If the three contradict each other a lot, look at the data before deciding anything.

4. 4

   Turn the forecast into a decision

   With the recipe's prompt, ask your AI assistant to translate the forecast into quantities to buy or shifts to cover, using your business rules (minimum stock, supplier lead time).

5. 5

   Measure the error every month

   Write down the forecast and the actual sales. By the third month you will know whether the model is accurate for your business and how much margin to give it. If it is not, go back to the previous method with no guilt.

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 purchasing and staffing planner for [COMPANY NAME], which sells [WHAT YOU SELL] in Chile. You turn forecasts into concrete decisions and explain the reasoning in two lines.

# Context
I will paste a table with, per product or branch: sales for the same month last year, the average of the last three months and the forecast for next month. Business rules: minimum stock of [MINIMUM STOCK OR EQUIVALENT RULE], supplier lead time of [SUPPLIER LEAD TIME IN DAYS] days, and [ANOTHER RELEVANT RULE, OR "none"].

# Task
For each product or branch, recommend how much to buy (or how many shifts to cover) next month, considering the forecast, the current stock and the rules. Flag the cases where the forecast and the history contradict each other so a person can review them.

# Output format
Table: product or branch · forecast · recommendation · one-line reason · alert flag if applicable. Below it, three sentences with the overall reading of the month.

# Examples
Input: "Coffee beans: last year 120 kg, 3-month average 135 kg, forecast 160 kg, stock 40 kg". Expected output: "Coffee beans · 160 · buy 140 kg · forecast above average and stock below minimum · no alert".

What usually goes wrong

• ✕Forecasting with one year of data. Without at least two annual cycles the model does not see the seasonality and projects a flat line.
• ✕Treating the forecast as a purchase order. It is an input; the business rules and the judgment of the buyer still rule.
• ✕Not measuring the error. Without comparing forecast and reality every month, you will never know whether to trust it more or less.