Quick Start
A forecast from an Excel workbook in about 5 minutes, then the same with your own data.
1. Open PowerForecast
Sign in. The sidebar holds Forecast Wizard, Results, file operations (upload / download a forecast specification), Reset Wizard, the documentation, the theme switch and the language.
2. Load the demo workbook
- On step 1 (Goal), click Start from an Excel workbook → Choose workbook…
- Pick
forecast-demos/04-us-retail-sales-budget/input/spec.budget.xlsx: 20 US retail series (sales by kind of business, millions of USD, monthly, January 2015 – December 2025). - The panel says Loaded spec.budget.xlsx: 20 series, 2,640 values; 9 settings from the Setup sheet. Every step is now filled.
3. Walk through the steps
| Step | What you see |
|---|---|
| 1 · Goal | Revenue selected (from the Setup sheet) |
| 2 · Your data | 20 series, 2,640 values, detected period Monthly, a preview of the first rows |
| 3 · How far ahead | 12 months. The Advanced section shows 95 %, baseline, Automatic |
| 4 · Review | the summary, the name US retail sales 2026 budget, an AI review of the settings |
4. Run
Click Run Forecast on step 4. A progress window shows the stages (checking your data → choosing a method → forecasting). With these 20 monthly series it takes about 6 minutes on a laptop (one core): each series gets several methods tested on its own past.
When it is done, See the results opens the Results page, and four JSON files (specification, results, diagnostics, profiling) are saved to your downloads so you can reopen the run later.
5. Read the result
- The verdict at the top, then the Overview: series forecast, typical error, better than a simple guess, the points to check, the chart.
- Pick a series in the chart's list (e.g. Online shopping), then Show per → Year to see yearly totals.
- Forecast table gives the numbers for 2026, month by month, with the likely range.
- Quality checks shows, per series, the method kept, its typical error and the other methods it beat.
The demo's scenario and expected results explain what to expect, including how the 2026 forecast compared with the real 2026 months.
6. Now with your data
Prepare a sheet like this (Excel or CSV):
| Date | Shoes | Bags | Belts |
|---|---|---|---|
| 2023-01-01 | 120 | 48 | 30 |
| 2023-02-01 | 135 | 51 | 28 |
| … |
- One row per period, the date in the first column (Excel dates are fine).
- One column per series, its name in the first row. A series may start later or stop earlier: leave those cells empty.
- For monthly data, two to three years is a good minimum. The more history, the better the seasons are learnt.
Then on step 2 drop the file, on step 3 say how far ahead, and run. To keep your settings with the data, use Save as workbook on step 1: it writes the Setup sheet and the Data sheet in one file.
Next: the Wizard Guide for every option, the User Guide for the Results page.