Planning Performance and KPI Interpretation

Planning Performance turns completed forecasts and operational documents into measured outcomes. Open Inventory > Niyu AI > Planning Performance or use the Inventory & Service, Forecast Quality, and Planning Impact tabs in Planning Overview.

Enable and build the baseline

1. Open Inventory > Configuration > Settings > Niyu Smart Stock.
2. Enable Performance Analytics for the company and save.
3. Continue normal forecasting, purchasing, transfers, receipts, and deliveries. The scheduled processing jobs collect durable facts and mature them over time.
4. Review the data-quality label before interpreting any KPI.

Performance pages do not recalculate years of history when the dashboard opens. They use stored evaluation records, delivery facts, inventory snapshots, and impact events so the dashboard remains responsive as data grows.

Data-quality labels

• Reliable: the measure has sufficient, mature source records for the selected period.
• Partial Coverage: the result is valid for the covered records, but part of the population lacks required evidence.
• Approximate: an allowed proxy or incomplete valuation basis was used. Read the formula and scope before acting.
• Collecting Baseline: the feature is active, but not enough time has passed to evaluate the period.
• Stockout-Censored: some demand periods were affected by unavailable stock and are separated from normal accuracy evaluation.
• Missing Cost Data: quantity evidence exists, but a dependable value calculation is not available.
• Insufficient Samples: too few mature observations exist for a dependable portfolio measure.

Unavailable is not the same as zero. A blank or unavailable KPI means the required evidence does not yet exist for the selected scope.

Inventory & Service

Customer service

• Unit Fill Rate compares units fulfilled by the original promise date with units ordered in the eligible delivery population.
• Customer On-Time Delivery compares deliveries completed by the original customer promise date with deliveries due in the period.
• OTIF counts a delivery only when it is both on time and in full against the original commitment.
• Operational Perfect Order Rate is optional. It adds the requirement that no recorded return exists during the configured return window.

Original promise dates are retained for evaluation. Moving a scheduled date later does not erase the original commitment or turn a late delivery into an on-time delivery.

Supplier performance

• Supplier On-Time Receipt compares eligible receipts completed by the original planned receipt date with receipts due in the period.
• Supplier Lead-Time Reliability summarizes the difference between actual receipt dates and original planned receipt dates. Positive delay values indicate late receipts.

Partial receipts are evaluated using the documented quantity and completion rules shown in the KPI information panel. Open the drill-down to reconcile the numerator and denominator.

Inventory health

• Inventory Turnover is value-based and appears only when dependable cost of goods sold and average inventory valuation are available.
• Unit Turnover is a separately labeled quantity measure when unit demand and average unit inventory overlap but dependable value history is unavailable.
• Stockout Rate reports observed product-warehouse days at or below zero stock for the eligible period.
• Excess Inventory Value and Dead-Stock Value use the configured planning and inactivity definitions together with available valuation data.

Forecast Quality

• WAPE is the primary portfolio accuracy KPI. It divides total absolute forecast error by total actual demand across eligible mature observations.
• Bias remains signed. Positive bias means over-forecasting; negative bias means under-forecasting.
• MAPE — Non-Zero Actuals excludes observations with zero actual demand and always shows its coverage.
• Forecast Confidence is a separate planning signal. It is never presented as realized forecast accuracy.
• Forecast Evaluation Coverage shows how much of the selected population has a mature, evaluable forecast snapshot.

A forecast is evaluated only after its horizon matures. Later forecast runs cannot overwrite the stored forecast used for that evaluation. Stockout-censored observations are identified separately so unavailable stock does not make a reasonable forecast look falsely high.

Planning Impact

• Action Adoption tracks recommendations accepted, overridden, ignored, or left pending, with their linked RFQs, purchase orders, and transfers.
• Purchase Spend Deferred measures purchase value delayed because an internal transfer covered the immediate need. It is not profit or permanent savings.
• Estimated Stockout Exposure and Estimated Stockout Value Protected use configured assumptions and clearly display their evidence quality.
• Exposed Value identifies eligible value that remained at risk after its action date or outcome window.
• Estimated ROI appears only when program cost, mature benefit evidence, configured assumptions, and sufficient samples are all available.

Verified operational facts remain separate from estimated financial impact. The impact ledger prevents the same outcome from being counted more than once across transfers, backorders, stockout protection, subsequent sales, and deferred purchases.

Filters, access, and reconciliation

• Use company, warehouse, date, vendor, product, category, and data-quality filters to compare like-for-like populations.
• Open a KPI card or its information control to see the formula, period, exclusions, maturity rule, numerator, denominator, and source status.
• Use the drill-down records to reconcile the displayed value. A KPI should not be accepted when its drill-down does not match its stated population.
• Financial values and cost-based measures are restricted to users with the appropriate manager and company access.

When a performance tab is empty

1. Confirm Performance Analytics is enabled for the current company.
2. Check that normal forecast runs and operational documents have occurred since it was enabled.
3. Read the label. Collecting Baseline and Insufficient Samples commonly require time, not reconfiguration.
4. Check company, warehouse, date, and data-quality filters.
5. Open the underlying Forecast Evaluation, Delivery Facts, Inventory Snapshots, or Impact Events menu when a manager needs record-level evidence.