Summary
The Default Leading Indicator Settings screen is where you configure how StockIQ uses leading indicators external metrics like CPI, vehicle registration, or a weather index to adjust statistical demand forecasts for a Demand Forecast Series.
It is accessible by clicking Admin > System Configuration > Leading Indicators > Default Leading Indicator Settings
Summary
Demand Forecast Series
Use the dropdown at the top of the screen to choose which Demand Forecast Series you'd like to configure. Each Demand Forecast Series has its own independent set of leading indicator settings, so be sure you've selected the right series before making changes.
General Settings
- Enable Leading Indicator Analysis for this Demand Forecast Series- Whether or not you'd like the for the whole Leading Indicator feature on this Demand Forecast Series. When off, no correlation analysis or forecast adjustment runs for any node/Custom Time Series under this Demand Forecast Series.
- Max Lead Periods (Monthly)- For monthly-interval Demand Forecast Series, the maximum number of periods the correlation search will lag a Custom Time Series ahead of demand when looking for the best-correlated offset (i.e.,how far back a leading indicator is allowed to "lead" demand).
- Max Lead Periods (Weekly)- Same as above, but used when the Demand Forecast Series's interval is Weekly instead of Monthly.
-
Max Ancestor Levels To Search- When a node has no usable direct correlation evidence of
its own (or a weaker one), the calculator walks up the hierarchy looking for a parent/ancestor with the same Custom Time Series and stronger (or any) evidence to borrow. This caps how many levels up it will walk before giving up -
Direct Mode Max Level Gap-If the Custom Time Series attaches at a hierarchy level that is this many levels
(or more) above the target node, "Direct" combination mode is suppressed and forced down to "Trend Adjust" instead — since a high-level indicator (e.g., national CPI) applied directly to a low-level node (e.g., item-site) produces nonsensical absolute values; it should only nudge the trend. - LI Trend Fade Periods- In Trend Adjust mode, once you move past the Custom Time Series' last observed value, the adjustment linearly fades from full effect down to zero over this many periods (rather than projecting the trend indefinitely).
Correlation Thresholds
-
Min Correlation To Apply- Minimum absolute Pearson correlation (|BestR|) required at the best
lag before an indicator is even considered "correlated enough" to apply. - Min Correlation For Window (Stability)- Minimum stability score (correlation strength calculated across a rolling sub-window, not just the full history) required to confirm the relationship isn't a fluke of the overall window
- Direct Combination R² Threshold- If the best R² meets/exceeds this threshold, the auto rule picks "Direct" combination mode (indicator drives the forecast value directly); otherwise it falls back to "Trend Adjust" (indicator only nudges the trend).
- Max Trend Adjustment Pct (fraction; 0.20 = 20%)- In Trend Adjust mode, caps how much the statistical forecast can be moved (as a fraction of its own value) by the leading-indicator adjustment, so one strong signal can't blow out the forecast.
History Requirements
- Min History Periods For Analysis- Minimum overlapping periods of demand and Custom Time Series history required before a correlation is even computed for a node/Custom Time Series pair.
- Min History Periods For Backtest- Minimum demand history length required to run the holdout backtest at all; also used to size the holdout window (roughly MinHistoryPeriodsForBacktest / 3, minimum 3 periods).
- Min History Periods For Segmented (Stability Window)- Window size used when computing the stability score — the detrended series is split into segments of at least this many periods each to check whether the correlation holds consistently across sub-windows, not just in aggregate.
- Backtest Min MAE Improvement Pct- Minimum percentage improvement in MAE (indicator-adjusted forecast vs. the StockIQ baseline forecast over the holdout window) required to keep the indicator applied.
Per-Custom-Time-Series Overrides
This grid lets you override behavior for individual Custom Time Series. Each row shows the Custom Time Series name, whether it's Enabled, and an optional Force Combination Mode (Auto, Direct, or Trend Adjust) if you want to lock in how that indicator is combined with the statistical forecast rather than letting StockIQ choose automatically. Unlike the rest of the screen, edits in this grid save immediately as you make them. There's no need to hit the page Save button for these rows.
Saving Your Changes
When you're finished adjusting the General, Threshold, and History settings, hit Save up top. You'll see a
confirmation dialog warning that modifying these settings can have a significant effect on how StockIQ calculates leading indicator adjustments. Confirm to persist your changes.
NOTE: These settings establish the default behavior for the selected Demand Forecast Series. Individual hierarchy
nodes can still be overridden independently. A node-level override always takes precedence over the defaults configured here, and (where configured) a node-and-indicator-specific override takes precedence over the per-Custom Time Series Force Combination Mode set on this screen.