Glossary term
Seasonal ARIMA (SARIMA)
Seasonal ARIMA, or SARIMA, is a time-series forecasting model that extends ARIMA by adding seasonal autoregressive, differencing, and moving-average terms.
Updated
Read time
What Is Seasonal ARIMA?
Seasonal ARIMA, or SARIMA, is a time-series forecasting model that extends ARIMA by adding seasonal autoregressive, differencing, and moving-average terms. It is used when data have both ordinary time dependence and a repeating seasonal pattern.
ARIMA models can handle trend and autocorrelation. SARIMA adds a seasonal structure, such as monthly data that repeat every 12 months or quarterly data that repeat every four quarters. That makes it useful for economic, retail, weather, inventory, traffic, claims, and demand forecasting.
Key Takeaways
- SARIMA is a seasonal extension of ARIMA.
- It models both nonseasonal and seasonal time-series dynamics.
- The model is often written as SARIMA(p,d,q)(P,D,Q)s.
- The seasonal period s depends on the data frequency, such as 12 for monthly data with annual seasonality.
- SARIMA can be powerful but depends on stationarity, diagnostics, and sensible seasonal structure.
How the Model Is Written
A common notation is SARIMA(p,d,q)(P,D,Q)s. The first three terms are the nonseasonal ARIMA components: autoregressive order p, differencing order d, and moving-average order q. The second group adds seasonal autoregressive order P, seasonal differencing order D, and seasonal moving-average order Q. The subscript s is the seasonal period.
For monthly sales with annual seasonality, s might be 12. For quarterly GDP-like data, s might be 4. The model tries to capture the relationship between current values, recent values, recent errors, and values or errors from prior seasonal periods.
Where SARIMA Is Useful
Use case | Seasonal pattern |
|---|---|
Retail sales | Holiday and monthly cycles |
Electricity demand | Weather and calendar cycles |
Hotel occupancy | Travel-season cycles |
Claims or call volume | Weekly, monthly, or annual repetition |
Economic indicators | Quarterly or monthly recurring patterns |
Financial Interpretation
SARIMA matters because many financial and operating decisions depend on forecasts that have seasonality. A business that mistakes a seasonal dip for a permanent decline may cut inventory or staffing too aggressively. A lender that misses seasonal cash-flow patterns may misread borrower risk. An investor comparing quarterly results may overreact if the seasonal baseline is wrong.
The model is most useful when the seasonality is stable enough to learn from history. If the business model changes, the customer base shifts, or a one-time shock disrupts the seasonal pattern, a SARIMA forecast can look precise while being economically stale.
Model Risk
SARIMA is not a magic forecasting engine. Analysts still need to check residuals, compare out-of-sample performance, test alternative specifications, account for holidays or calendar effects, and understand whether external drivers matter. A purely historical model can struggle when a new policy, competitor, technology, or supply shock changes the process.
SARIMA can also be confused with seasonal adjustment. Seasonal adjustment removes or estimates seasonal patterns to make data easier to compare over time. SARIMA is a modeling framework that can be used in forecasting and, in some systems, as part of seasonal adjustment workflows.
For finance teams, the model is usually strongest when paired with business judgment. A model can see a December sales pattern, but it may not know that a product launch moved to November, a promotion was canceled, or a supplier constraint capped volume.
Analysts should also separate forecast accuracy from explanatory power. A SARIMA model may forecast a seasonal series well without explaining why the pattern exists. For budgeting and risk work, that means the model output should be paired with operating knowledge about customers, capacity, pricing, and calendar effects.
The Bottom Line
Seasonal ARIMA is an ARIMA-style forecasting model with explicit seasonal terms. It is useful when data repeat in predictable cycles, but its financial value depends on good diagnostics, stable seasonality, and judgment about whether the past remains a useful guide.