Build a rolling 12-month sales forecast with ARIMA in Excel

Step 1: Prepare the monthly historyExcel

Lay out at least 24-36 months of monthly sales in a single clean column with dates, no gaps or merged cells — ARIMA needs a continuous series to fit properly.

Step 2: Ask Copilot to build the modelCopilotExcel

Use Excel's Python integration through Copilot instead of a manual trendline, so the model accounts for seasonality and autocorrelation.

Prompt idea:

Using the monthly sales data in column B, write a Python ARIMA model in this sheet to forecast the next 6 months. Show the model's order parameters, plot actuals vs. forecast, and flag if the series isn't stationary.

Step 3: Sanity-check the model choiceCopilot

Ask Copilot to explain why it picked those ARIMA parameters (p, d, q) and what the confidence interval looks like — a forecast without an explainable basis isn't one you can defend in a planning meeting.

Step 4: Chart and shareExcel

Turn the forecast into a chart with the confidence band shown, and note the model's assumptions directly on the sheet so anyone revisiting it in three months knows what changed if actuals diverge.

Check the work

• Confirm the historical data has no missing months before trusting the fit.
• Ask Copilot to explain the ARIMA order in plain language — if it can't, don't ship the forecast.
• Compare the model's backtested predictions against the last 3 known months before extending it forward.

Source: Nexacu, "Copilot for Excel (2026): Practical Prompts, Agent Mode & Time-Saving Examples" (2026)