Evaluate a statistical forecast's accuracy and explain the misses
A backtest and error review of a demand or revenue forecast against simple benchmarks, showing where it misses and how to improve it.
The prompt
Help me evaluate how accurate a forecast has been and explain where and why it missed.
What is forecast: [e.g. "weekly demand by SKU", "monthly revenue by region"]
Forecast method: [e.g. moving average, ARIMA, Prophet, ML model, manual planner adjustments]
Data: [PASTE forecast vs. actuals by period and segment, with the forecast date and horizon]
How the forecast is used: [e.g. "inventory orders placed 6 weeks ahead"]
Steps:
1. Choose error metrics that fit the use and explain the choice: MAE, MAPE (and why it breaks near zero), weighted MAPE, RMSE, and bias (systematic over- or under-forecasting).
2. Calculate or show me how to calculate error by horizon, by segment, and over time.
3. Compare against simple benchmarks: naive (last value), seasonal naive, and moving average. A forecast that cannot beat these needs rethinking.
4. Find patterns in the misses: specific segments, promotions, holidays, new products, structural breaks, data issues.
5. Separate model error from the effect of manual overrides, if any.
6. If the method is being changed, describe a proper backtest with rolling-origin evaluation.
Return:
- An accuracy summary table: Segment | Horizon | Error metric | Bias | vs. benchmark.
- The 3-5 biggest sources of error, with evidence.
- Specific improvements ranked by expected impact and effort.
- A short summary for business stakeholders on how much to trust the forecast and the right size of safety buffer.
Use only the numbers I provide. Mark anything I need to supply as [ADD DATA].
Highlighted [brackets] are blanks to replace with your own details.
Fill in the blanks and use this promptMore Data & Analysis prompts
What to fill in
- e.g. "weekly demand by SKU", "monthly revenue by region"
- e.g. "inventory orders placed 6 weeks ahead"
- ADD DATA
How this prompt is built
This prompt follows the Few-Shot structure (Few-shot prompting). A few-shot prompt includes one or more worked examples of the input and the output you want, so the model can copy the pattern instead of guessing it.
Related prompts
- Solve an Excel or Google Sheets problemGets you the right formula, pivot, or approach — with an explanation you can reuse.
- Find missing values in my data and explain their impactProfiles blanks and placeholder values by column, spots patterns in what's missing, and recommends how to handle each gap.
- Explain a dataset or report in plain business languageExplains what a dataset or report contains, what each column or metric means, and what it shows, in words anyone can follow.
- Break down why a number changed between two periodsA driver breakdown and bridge that shows exactly which segments, volume, price, or mix effects caused a metric to move.