Subject: DEMO, monthly demand forecastModel: monthly demand forecastGenerated: in your browser, nothing uploaded
The one question this report answers: does your forecast actually beat a fair naive benchmark, out-of-sample? Scored against the toughest of three naive baselines (persistence, climatology, and a seasonal naive when the data is seasonal) using standard published scoring rules.
Result
Periods scored (out-of-sample)
24
Benchmark (toughest naive)
random-walk (persistence)
Your forecast error (MAE)
9.817
Naive-benchmark error (MAE)
6.983
Forecast Value Added vs naive
−40.6%
MASE (<1 beats naive)
1.406
Beat-the-naive rate
33% of periods
Bias (mean over/under)
−0.925
FAILThis forecast does not beat the naive benchmark. The simple baseline is as good or better, out-of-sample.
How to read it: a higher-effort forecast that loses to a naive baseline is adding cost without adding accuracy. The honest move is to ship the naive baseline (or fix the model) until it can clear the bar. A FAIL is a valid, valuable finding, often the most valuable one.