Summary
The Leading Indicator Details Dialog shows the correlation analysis behind any leading indicators configured for an item, and explains how (or whether) each one is being used to adjust the item's forecast.
It is accessible from Forecast > Forecast Manager by selecting an item and clicking Edit > Leading Indicator Detail
Leading Indicators Grid
Column Description
- Custom Time Series - Name of the leading indicator (e.g. "Point-Of-Sale")
- Applied From - Which hierarchy level the correlation was detected at, and where it's being used
- Best Lag - The lag period where correlation was strongest (e.g. "3"); tells you how far ahead the leading indicator predicts
- Pearson r (Detrended) - Correlation strength with trend removed; the primary measure of signal quality
- Pearson r (Raw) - Correlation strength on raw data, trend included
- R² - Coefficient of determination; the % of variance explained (e.g. 0.985 = 98.5%)
- Stability - Consistency of the correlation across backtest windows; higher is more reliable
- Scale Factor - How to convert the leading indicator's units into demand units
- Mode Used - Which combination method was used (Auto, Direct, etc.)
- Not Applied Reason - If the indicator isn't being used, why (see below); blank or "Auto" means it's active
- Backtest MAE Baseline - Average forecast error without the leading indicator
- Backtest MAE With LI - Average forecast error with the leading indicator applied
- Backtest Improvement % - How much better the forecast got (e.g. "15% better"); large and positive means the indicator is genuinely helping
- Date Updated - When this analysis was last run
Correlation Detail
Clicking a row in the grid opens the full analysis for that leading indicator.
Left side - Scatter Plot: plots the leading indicator's values (X-axis) against actual demand (Y-axis), with a red
regression line showing the best fit. Points clustered tightly around the line indicate strong correlation; scattered
points indicate a weak one.
Right side - Info/Statistics: repeats Applied From, Combination Mode Used, and Applied (Yes/No) from the table above, plus the same correlation and backtest statistics (Pearson r Detrended/Raw, R², Stability Score, Best Lag, Scale Factor, Backtest MAE Baseline/With LI, Backtest Improvement %).
Why a Leading Indicator Might Not Be Applied
StockIQ runs backtests and stability checks before applying any leading indicator. If it isn't applied, the reason
appears in Not Applied Reason:
| Reason | What It Means | How to Fix |
| Insufficient History | Not enough data points to detect correlation reliably | Wait for more historical data to accumulate |
| Backtest Failed | The correlation didn't improve forecasts in historical windows | The signal may be random or unreliable; review the data |
| Unstable Correlation | Correlation strength varies wildly across stability backtest windows | The relationship isn't consistent; may not be a true signal |
| Correlation Below Threshold | The relationship isn't strong enough to trust | Lower the Min Correlation To Apply threshold, or find a different leading indicator |
Stability scoring guards against coincidental ("blind squirrel") correlations by validating across multiple time
windows (controlled by Min History Periods For Segmented) and confirming the correlation's direction holds
consistently. If it reverses or disappears between windows, the indicator is flagged Unstable Correlation and won't be applied — controlled by the Min Correlation For Window (Stability) threshold.
Limitations On Leading Indicators
Leading indicators can't predict unforeseen shocks, compensate for bad or incomplete source data, work reliably on short histories, or bridge structural breaks in the business (new product, new market, new customer). They also require consistent, complete, correctly-aligned, and unit-consistent data to produce a usable correlation.
Leading Indicator FAQs
How long before a leading indicator starts working? - At least 2-3 full cycles of the lag period — for a 3-month lag, that's 6-9 months of data minimum.
Can I have multiple leading indicators for one item? - Yes. StockIQ tests and applies them independently, merging their effects into the final forecast.
What if the leading indicator stops correlating? - The next backtest will fail and StockIQ will stop applying it, shown under Not Applied Reason.
Can I adjust the lag manually? - No, StockIQ detects it automatically. You can constrain the search range with Max Lead Periods (Monthly or Weekly).
Does StockIQ use the leading indicator for safety stock? Yes — safety stock adjustments follow automatically when the forecast changes.
What does Combination Mode do? It controls how the leading indicator is mathematically merged with the base forecast (Auto lets StockIQ choose; Direct uses direct combination with configurable thresholds).
What if my leading indicator's units don't match demand? The Scale Factor handles this — StockIQ scales the leading indicator to match demand units and shows the scaled version in the chart.
What's the difference between Pearson r (Raw) and Pearson r (Detrended)? Raw includes the trend of both series; Detrended removes the trend first, showing the pure correlation in the fluctuations. Detrended is usually the cleaner signal measure.