← All analytics projects
02
Project 02 · Time-series planning

U.S. Retail Sales Forecast

Estimate a plausible national retail-sales range, validate it against recent holdout data, and compare it with a simple year-ago baseline.

Time series36.4% MAE gainBacktesting
Latest observationJul 2026Committed snapshot
Latest sales$784.6BMillions USD converted to billions
Backtest MAE$18.1BLatest 24-month holdout
Gain vs baseline36.4%Lower MAE than year-ago
Planning workspace

Explore a 12-month forecast

The backtest remains fixed while the selected future horizon changes, keeping comparisons honest.

Forecast horizon

$628.9B$773.6B$918.4BAug 2025Jul 2026Jul 2027
ObservedHolt–Winters forecastApprox. 80% band
Validation

Compare against a credible seasonal baseline

Latest 24-month holdout

$607.4B$719.7B$831.9BAug 2024Jul 2025Jul 2026
ActualHolt–WintersYear-ago baseline
Model MAE$18,145M
Model RMSE$21,869M
Seasonal-naive MAE$28,509M
Seasonality

Recurring calendar pattern retained

Average monthly sales index over the latest 10 years; each year's average is 1.00.

Jan0.89
Feb0.87
Mar1.00
Apr0.97
May1.04
Jun1.02
Jul1.02
Aug1.03
Sep0.98
Oct1.01
Nov1.03
Dec1.13