Combining seasonality and recent demand signals into an out-of-sample weekly forecasting system
Integrated model · provisionally validated
Expected seasonal demand provides context for allocating advertising investment throughout the year. Use the baseline alongside acquisition efficiency, campaign objectives, and service capacity when deciding where to accelerate or temper spend.
Anticipated demand periods can inform mail-drop schedules and related promotions. Account for production and delivery lead times, then coordinate campaign timing with the expected seasonal ramp.
Compare observed acquisition with forecasted demand to distinguish normal seasonal increases from unusually strong or weak performance. Investigate gaps alongside campaign activity, conversion, capacity, and tracking changes.
These are practical decision-support applications of a provisionally validated forecast. The 19.2% lower holdout MAE and 12.7% lower holdout RMSE measure predictive performance; they do not establish deployed budget-optimization outcomes, incremental marketing lift, or causal attribution. Forecast gaps alone cannot isolate a marketing effect.
Can an interpretable adaptive seasonal model outperform simple benchmarks on genuinely unseen demand?
Fixed seasonality misses annual level changes; flat recent averages miss seasonal direction. Integrating shape, relative level, and local momentum addresses both weaknesses.
2023–2024 development used chronological rolling-origin selection. The selected architecture and score scale were frozen before evaluation on the untouched 2025 holdout. The final model uses no weather or economic inputs.
S(w) is the retained two-harmonic seasonal factor. L is the documented six-week endpoint trend of completed normalized observations. Each forecast excludes the target week.
MAE · weekly New Customers · lower is better
RMSE · weekly New Customers · lower is better
2025 holdout supporting diagnostics: predictive R² = 0.778; actual–predicted correlation = 0.891. Forecast error relative to the recent four-week benchmark is the primary evaluation.
Improvements are the source-reported values calculated before rounding; displayed two-decimal errors do not reproduce those percentages exactly.
The supplied sources report holdout summaries but do not contain all 52 actual/predicted 2025 rows. A weekly holdout line chart cannot be reconstructed accurately from these summaries.
Provisionally validated: the model generalized usefully to one untouched annual holdout. The later 2026 check is a separate, approximate alignment-sensitivity analysis.
Assumes uniform within-week timing. Exact validation needs daily Customer Since dates or daily counts.
Calendar definitions are part of the model specification, not a formatting detail.
MAE · weekly New Customers · approximate calendar alignment
The model’s advantage persisted under this additional alignment check, but the margin was smaller: approximately 1.56 versus 1.62 New Customers per week in MAE. This is not a repeat of the untouched 2025 holdout. The next checkpoint is exact daily re-bucketing of 2026 New Customers, followed by a frozen-model rerun without retraining.
One annual holdout supports provisional rather than full validation. The 2026 approximation assumes uniform timing within each pivot week; exact daily counts are required for a definitive comparison.
Flagship integration: seasonality, adaptive demand level, and local momentum form the final architecture. Weather and economic candidates were tested but not retained. The next layer translates forecasts into a Demand Score.
Model structure, calculated validation metrics, and limitations are documented in this series. Source workbooks and full methodology downloads containing underlying business observations are withheld.
Trent Turner