PowerForecast in plain words
Everything a planner needs to make a forecast and judge it. No statistics required.
What PowerForecast does
You give it the past: how much you sold, served, received or spent, period by period (day, week, month…). It gives you the future: the most likely value for each coming period, with a likely range (low – high) around it, and a verdict that says whether you can plan with it.
It works the way a careful analyst would:
- It studies each series: does it grow? does it repeat every year? are there odd months?
- It tests several methods on your own past. It hides the most recent periods, forecasts them from what came before, and compares with what really happened. The method that would have been the most accurate is kept.
- It tells you how far off it typically was in that test, and whether it beat the simplest possible guess (repeating the last value). A method that cannot beat that guess is not worth trusting, and the page says so.
The 4 steps of the wizard
| Step | You do | Tip |
|---|---|---|
| 1 · Goal | Say what you forecast (demand, workload, revenue…), or start from an Excel workbook | A workbook can fill everything at once |
| 2 · Your data | Drop your Excel or CSV file: the date in the first column, one column per product, store or department | The time unit (monthly, weekly…) is read from your dates |
| 3 · How far ahead | Type a number of periods, or click 3, 6, 12, 24 months… | The further ahead, the less certain. Forecast what you need to plan |
| 4 · Review | Check the summary, give the forecast a name, Run Forecast | An AI reviews the settings before you run |
Expert settings (time unit, width of the likely range, scenario, method) are in a closed Advanced section. The automatic choices suit most forecasts.
Reading the result
The Results page opens with a verdict:
| Verdict | Means | Do |
|---|---|---|
| Ready to use (green) | Every series has a forecast that beat the simple guess, with a modest typical error | Plan with it |
| Usable · n to check (orange) | The forecast exists, but some points deserve a look | Read the points to check before you rely on the numbers |
| Not usable yet (red) | Some series got no forecast | Check the data step (usually too short or irregular history) and run again |
Under the verdict:
- Typical error: how far off the forecast was, on average, when tested on your past (in %). Below 10 % is good for most planning. Above 20 %, use the numbers as a direction, not as exact values.
- Better than a simple guess: in how many series the method beat repeating the last value.
- Points to check: one sentence each, for example "The forecast tends to be too low" (consider planning a little higher) or "Not better than a simple guess" (treat with care, more history usually helps).
- The chart: what happened (grey) and the forecast (coloured), with the likely range around it. Show per adds up periods into weeks, months, quarters or years.
The other tabs go deeper: Forecast table (the numbers, to copy into your plan), Your data (what was found in your history: trend, seasons, quality), Quality checks (the test results, the methods compared, an AI critic that can propose and try a change).
When a point needs checking
| The page says | Usual cause | What to try |
|---|---|---|
| Not better than a simple guess | short or very irregular history | add more history; forecast fewer periods ahead |
| Large typical error | volatile series (promotions, one-off events) | use it as a range; split very different products into separate series |
| Tends to be too low / too high | a recent change in level the method follows slowly | plan a little above / below; re-forecast when new actuals arrive |
| Data quality is fair or poor | gaps, duplicates, outliers | fix the data in Excel and run again |
Words used on the screen
| Word | Means |
|---|---|
| Series | one line of history: a product, a store, a department |
| Period | one step of time: a day, a week, a month… |
| How far ahead (horizon) | how many periods are forecast |
| Likely range | the band where the real value should fall (95 % = 19 times out of 20) |
| Typical error | the average gap between forecast and reality on your past, in % |
| Simple guess | repeating the last value: the minimum a method must beat |
Next: Quick Start, then the Wizard Guide.