Forecast Review
Judges on the error distribution rather than on the last data point, which is noise.
Free — no card needed
Create a free account
to open Forecast Review — and the other 364 prompts across 21 categories.
Already have an account?
Running it, start to finish
- Look at the whole error distribution, not the latest period.
- Compare against a naive baseline before doing anything else.
- Establish which direction of error actually costs more.
What you get back
The output this produces, every time.
- Judges on the error distribution rather than on the most recent period.
- Compares against a naive baseline, which decides whether the model earns its complexity.
- Checks the cost asymmetry, since minimising symmetric error optimises the wrong thing.
Getting better results
Where this usually goes wrong, and how to avoid it.
- Always run the naive baseline. A model that does not beat last period's actual is complexity with no benefit, and this check takes minutes.
- Watch for repeated tuning against one holdout. It gradually fits the test set, and the over-fitting is invisible because each individual adjustment seemed reasonable.
- Consider responsiveness instead. Where the process is genuinely unpredictable, shorter lead times beat any further forecasting effort.
More Data Analysis prompts
All Data Analysis →- Data Analysis
Segmentation Analysis
Produces segments that are actionable and stable, since statistically clean but unusable is the standard failure.
ClaudeChatGPT002Open → - Data Analysis
Metric Definition Review
Treats two teams reporting one metric differently as a governance problem rather than a query bug.
ClaudeChatGPT000Open → - Data Analysis
A/B Test Design and Reading
Designs a test that can actually detect the effect you care about, and states the decision rule before the data arrives.
ClaudeChatGPT000Open →
Written for The AI University. Every prompt in this library is original work — authored, tested and revised here, not collected from elsewhere. 365 of them, free with an account.