On-Time In-Full (OTIF)

OTIF is a logical AND gate, not an average. A delivery fails the metric if it is late but complete, or on-time but short-shipped. This strict binary evaluation prevents suppliers from masking poor performance in one dimension with excellence in the other.

• On-Time: Delivery occurs on or before the original requested date, not a renegotiated date.
• In-Full: The quantity received exactly matches the quantity ordered; partial shipments are failures.
• Composite Rate: Calculated as (Number of OTIF-Compliant Lines / Total Lines Ordered) × 100.

A critical distinction in OTIF calculation is the anchor date. High-maturity supply chains measure against the customer's original request date, while less mature ones measure against the supplier's acknowledged date.

• Customer Request Date: The date the buyer initially wanted delivery. Measuring against this exposes the full capability gap.
• Supplier Commit Date: The date the supplier promised after order acceptance. This allows suppliers to negotiate away poor performance.
• Best Practice: World-class organizations use the unadjusted customer request date to drive genuine lead time compression.

Traditional fill rate metrics often obscure true performance by allowing backorders or partial shipments to be counted as fulfilled. OTIF eliminates these loopholes.

• Fill Rate: Measures the percentage of demand met from available stock, often allowing late deliveries to count positively.
• OTIF: Imposes a strict time window; a perfect fill delivered one day late is a failure.
• Line-Level vs. Order-Level: OTIF is typically measured at the line-item level. Order-level OTIF, where every line on a purchase order must be perfect, is a more stringent variant used for critical assemblies.

Accurate OTIF measurement requires clean, synchronized data from both the buyer's procurement system and the supplier's shipping documentation.

• Required Data Fields: Purchase order line number, ordered quantity, original request date, actual receipt date, and actual receipt quantity.
• Data Quality Challenges: Discrepancies in goods receipt posting times, time zone mismatches, and unrecorded partial deliveries corrupt the metric.
• Automation: Leading organizations integrate ERP systems with supplier portals to capture real-time ASN data, eliminating manual data entry errors.

Procurement organizations use rolling OTIF scores to classify suppliers into performance tiers, which directly informs sourcing decisions and inventory strategies.

• Strategic Partners: >95% OTIF — eligible for long-term contracts and reduced oversight.
• Core Suppliers: 85-95% OTIF — require collaborative improvement plans.
• Transactional/At-Risk: <85% OTIF — trigger formal corrective action requests and potential resourcing.
• Dynamic Safety Stock: Suppliers with high OTIF variability force buyers to hold larger buffer inventories, directly increasing carrying costs.

OTIF is a lagging indicator of historical performance. Predictive lead time analytics transforms this reactive metric into a forward-looking operational capability.

• Lead Time Prediction: Machine learning models forecast the probability of an OTIF failure for each open purchase order before the delivery date.
• Prescriptive Intervention: When a model predicts a high risk of failure, the system can trigger expediting actions or re-route inventory from alternative nodes.
• Root Cause Analysis: Correlating OTIF failures with external variables—such as port congestion or supplier financial distress—enables proactive risk mitigation rather than post-mortem reporting.

The application of machine learning models to forecast the total elapsed time from purchase order issuance to goods receipt. Unlike static lead times in ERP systems, these models dynamically account for:

• Supplier processing variability
• Manufacturing cycle times
• Transit duration and port congestion
• Seasonal demand patterns

Accurate lead time prediction is the foundational input for setting realistic delivery dates and improving OTIF scores.

A composite quantitative metric that ranks suppliers based on historical delivery precision, lead time stability, and responsiveness. Key components include:

• On-Time Percentage: Frequency of meeting committed dates
• In-Full Percentage: Frequency of delivering complete quantities
• Lead Time Variance: Consistency of delivery cycles
• Recovery Rate: Speed of resolving disruptions

These scores enable procurement teams to segment suppliers by risk tier and trigger corrective action plans for chronic underperformers.

A range of values within which a future delivery date is expected to fall with a specified probability (e.g., 90%). Unlike a single point estimate, prediction intervals quantify uncertainty and enable:

• Dynamic buffer time calculations for safety stock
• Risk-adjusted order promising to customers
• Proactive exception alerts when confidence drops

Prediction intervals transform OTIF from a reactive lagging indicator into a forward-looking risk management tool.

An algorithmically adjusted time cushion added to a deterministic lead time forecast based on real-time risk factors:

• Current port congestion indices
• Supplier production backlog signals
• Weather disruption probabilities
• Geopolitical risk scores

Dynamic buffers replace static safety lead times, absorbing variability precisely when and where it is needed to protect OTIF performance without inflating inventory costs across the board.

A scale-independent accuracy metric measuring the average absolute percentage difference between forecasted and actual lead times. While widely used for benchmarking model performance, MAPE has limitations:

• Asymmetric penalty: underestimates are penalized differently than overestimates
• Undefined when actual values are zero
• Sensitive to outliers in sparse delivery data

Supply chain data scientists often complement MAPE with pinball loss functions and prediction interval coverage probability for a more complete accuracy assessment.

A set of methods and techniques that make complex machine learning predictions understandable to supply chain planners. When a model predicts a delivery will miss OTIF targets, XAI reveals the key drivers:

• SHAP Values: Quantify each feature's contribution to a specific delay prediction
• Counterfactual Explanations: Show what would need to change to achieve on-time delivery
• Feature Importance Rankings: Identify systemic issues like a specific carrier lane or product category

This transparency builds planner trust and enables targeted corrective actions rather than black-box alerts.