Demand Forecasting · Market Analysis
US Retail & Food Services Sales
18-Month Forward Forecast
Holt-Winters triple exponential smoothing applied to 135 months of seasonally adjusted retail sales data from the Federal Reserve Bank of St. Louis. The model projects $789.2B in March 2027 (+4.9% from current), with a 95% confidence interval of $705B–$873B. Forecast uncertainty widens with horizon: the 80% band captures plausible outcomes under stable macro conditions; the 95% band accounts for demand shocks comparable to those observed in 2020–2021.
Latest Actual
$752.1B
March 2026 · RSAFS
12-Month Forecast
$789.2B
March 2027 · point estimate
Projected Change
+4.9%
vs. March 2026 actual
Model RMSE
$12.4B
in-sample · 1.65% of mean
Historical Sales + 18-Month Forecast with Confidence Intervals
Gray: model fit on historical data · Amber: point forecast · Bands: 80% and 95% predictive intervals
Year-over-Year Growth Rate
COVID-19 shock: −19.7% (Apr 2020) · Recovery peak: +51.9% (Apr 2021)
Model Fit: Actual vs Fitted Values
In-sample RMSE $12.4B · 1.65% of mean · AIC 2,577
// Methodology & Assumptions
Data
FRED series RSAFS — Advance Retail and Food Services Sales, Seasonally Adjusted, monthly, millions of USD.
135 observations from Jan 2015 through Mar 2026.
Source →
Model
Holt-Winters triple exponential smoothing with additive trend and additive seasonal components
(statsmodels.tsa.holtwinters.ExponentialSmoothing, seasonal_periods=12).
Parameters optimized by maximum likelihood. AIC: 2,577. Smoothing level α=1.0.
Uncertainty
Predictive intervals derived from in-sample residual standard deviation ($12.4B) scaled by
√h where h is forecast horizon. 80% CI uses z=1.28; 95% CI uses z=1.96.
Intervals assume homoscedastic residuals and do not capture structural breaks.