Apply Seasonal Adjustments Consistently Across the Workbench

Haley Vicini

1. Abstract

Consistent seasonal treatment across all indicators in a Workbench ensures meaningful comparisons and prevents misleading correlations caused by mixing adjusted and non-adjusted series. Applying seasonal adjustments intentionally and uniformly strengthens both exploratory analysis and downstream modeling.

2. Context

Apply this best practice when building or reviewing a Workbench that includes time-series data with known seasonal patterns. This is especially important in industries such as retail, hospitality, manufacturing, and energy, where recurring cycles significantly influence performance.

3. Content

3.1 Why It Matters

Seasonality is a recurring pattern tied to calendar effects, weather cycles, fiscal schedules, or consumer behavior. When one indicator is seasonally adjusted (SA) and another is not (NSA), their patterns may appear misaligned even if they are economically related.

Inconsistent seasonal treatment can:

• Create artificial lag relationships
• Distort correlation strength
• Mislead interpretation of turning points
• Introduce instability into Predict models

For example, an unadjusted retail sales series that spikes every December will not align cleanly with an employment series that has already been seasonally adjusted. The mismatch may appear as a weak relationship or a false lead.

The goal is not to adjust everything blindly, but to ensure that indicators are treated consistently and intentionally.

3.2 How to Apply

When preparing a Workbench:

1. Confirm the seasonal status of each indicator.
   Check whether the source data is already seasonally adjusted using the seasonality flag in the workbench.
2. Apply seasonal adjustment where appropriate.
   If the data is not seasonally adjusted, use the Transform Modal to apply Seasonal Adjustment before analysis or use the Bulk Transformation feature to apply Seasonal Adjustment to all indicators with seasonality present.
3. Maintain consistency across related indicators.
   If comparing or modeling two economically linked series, ensure both are either SA or consistently treated.
4. Re-run Discover diagnostics.
   After adjusting, verify that correlations and lead relationships behave more intuitively.
5. Document the seasonal treatment choice.
   Note whether indicators are SA or NSA for transparency and future review.

3.3 Example

A Workbench includes retail sales (NSA) and unemployment (SA). Initial Discover results show inconsistent timing between the two. After applying seasonal adjustment to retail sales, correlations strengthen and lead relationships become more intuitive, improving analytical clarity.

3.4 Common Pitfalls

• Mixing SA and NSA indicators without realizing it
• Applying adjustment inconsistently across similar indicators
• Confusing structural breaks (e.g., post-pandemic shifts) with seasonal effects

3.5 Expected Results

• Clearer trend alignment across indicators
• Stronger and more interpretable correlations
• More stable coefficient behavior in Predict
• Improved confidence in both exploratory analysis and forecasting