The supply chain director walks into the quarterly review with what looks like a clean OTIF number. Ninety-one percent. Then the largest customer of the company sends over a vendor scorecard that puts the same operation at seventy-six percent on the same period. The numbers do not reconcile. The internal team measured against the date a load left the dock. The customer measured against the date the goods landed in their warehouse. The internal team counted a delivery as in-full when the line items matched the picked quantities. The customer counted in-full against the order they placed, including the units that were short-shipped and back-ordered. Same operation, two networks of measurement, fifteen-point gap.
This is the lived reality of OTIF in a multi-site manufacturer. The metric sounds simple, but the measurement surface is full of edge cases that bend the number in directions that benefit the side doing the measuring. For manufacturers running more than one production or distribution location, the problem compounds because a single customer order may be sourced from multiple sites, fulfilled across multiple shipments, and confirmed at multiple receipt windows. Without disciplined measurement, the OTIF number becomes a story the operation tells itself, while the customer scorecard tells a different story that determines whether the contract is renewed.
Why On-Time and In-Full Are Different Animals
The first step in honest multi-site OTIF tracking is to separate the two halves of the metric. On-time and in-full are not the same thing, they are not driven by the same process failures, and they do not respond to the same fixes. Treating them as a single combined number obscures the diagnostic signal that tells the operation what to do next.
On-time is a logistics metric. It measures whether the goods reached the customer by the agreed date. The drivers of on-time performance are dispatch scheduling, carrier reliability, transit time variability, and the time between confirmation and load-out. A network that is failing on-time is usually failing somewhere in the post-production phase, after the materials and finished goods exist but before they reach the customer. The interventions are operational and often involve carrier management, dock scheduling, and load planning.
In-full is an inventory and production metric. It measures whether the right quantity was actually shipped against the order. The drivers of in-full performance are demand visibility, available-to-promise discipline, production execution against schedule, and the accuracy of stock data at the moment of allocation. A network that is failing in-full is failing earlier in the chain, before the goods are even loaded. The interventions are upstream and involve planning, allocation logic, and the integrity of inventory records.
Multi-site manufacturers routinely excel on one of these and struggle with the other, and which one they struggle with is diagnostic. A network with strong on-time and weak in-full has a planning problem. A network with strong in-full and weak on-time has a logistics problem. A network with weak on both has a coordination problem between the two. The on time in full kpi is a starting point. The decomposition is the actionable signal.
How Multi-Site Networks Distort the Measurement
The complications begin when the order does not ship from a single site. Customer orders in multi-site networks are often allocated across two or three locations based on stock availability, geographic proximity, or production capacity. Each shipment has its own dispatch date, its own carrier, its own transit time, and its own receipt window. The customer sees one order. The internal system sees three loads. The OTIF calculation has to choose how to count.
The two common choices both have problems. Counting per-shipment inflates the metric because each shipment is treated as an independent observation, and a customer with one bad shipment out of three still gets credited with two on-time observations. Counting per-order is honest, but it requires that every shipment for the order arrive in the agreed window, which means a single late shipment from one site fails the entire order. Most internal teams default to per-shipment because the number looks better. Most customer scorecards default to per-order because the customer experiences one delivery per order, not three.
The right answer for multi-site OTIF tracking is to maintain both views, with per-order as the headline and per-shipment as the diagnostic. The per-order number tells the operation what the customer sees. The per-shipment number tells the operation which sites are causing the misses. Without both, the team optimizes the wrong thing.
There is also the reservation problem. When the same stock looks available to two orders at the same moment, the system commits to both, then fails to deliver one of them in-full. This happens whenever inventory is visible across the network without a state machine that locks reservations as they happen. The fix is structural. The transfer engine and the order allocation logic must share a single source of truth for what is actually available, and that source must reflect reservations the moment they are made.
The ATP Layer That Keeps the Metric Honest
The OTIF problem starts at the commit. When a customer asks for a quantity by a date, the operation either has a credible plan to deliver or it does not. Available-to-promise is the calculation that distinguishes between the two. ATP at order entry tells the system what can be honestly committed given current stock, reserved quantities, incoming receipts, and production schedules across every site in the network. An order taken against an honest ATP value has a structural chance of being on time and in full. An order taken against a stock number that ignores reservations and incoming demand is failing the OTIF test at the moment it is accepted.
Multi-site networks add a wrinkle to ATP because availability is not a single number. The same item may be plentiful at one site and short at another, and whether it is available to the customer depends on the network's allocation rules. A serious ATP implementation handles this by computing per-site availability and per-network availability, with the allocation rule explicit rather than implicit. This lets the order entry team see whether a commit can be honored from a single site, whether it requires a transfer between sites, and whether it depends on incoming production. Each of those answers carries a different OTIF risk profile.
For more on how this calculation looks on the factory floor, the available-to-promise piece in the FalOrb archive walks through the bottleneck logic and the per-product view. The same logic applies to multi-site OTIF, with the additional layer of network allocation on top.
At-Risk Order Alerts and the Window to Recover
Once the order is committed, the second discipline that protects OTIF is early warning. An order that is going to miss its date or its quantity rarely fails at the last moment. The signals appear days in advance. A production run is behind schedule. A transfer is stuck in pending. A receipt is overdue. A reservation has been bumped by a higher-priority order. Each of these is a leading indicator that the OTIF outcome is at risk, and each is recoverable if the team learns about it in time.
A network without alerting on at-risk orders only learns about misses on the day the customer expected the delivery. By then the recovery options are narrow and expensive. A network with proactive alerts on at-risk orders has days or weeks of warning, and the recovery options include alternative sourcing, expedited transfers, partial dispatches with back-ordered remainders that the customer accepts, and renegotiated dates that the customer agrees to in advance. The metric improves not because the underlying performance is better but because the recovery cycle is faster.
The platform requirement is straightforward. Alerts must be tied to specific orders, specific risks, and specific recovery actions. A general dashboard that shows the OTIF percentage trending down does not help anyone. An alert that names the order, the customer, the at-risk quantity, the bottleneck, and the proposed action gives the operations team something to do. Multi-site networks need this even more than single-site operations because the recovery action often involves coordination between sites that would otherwise not be talking to each other.
The Customer Scorecard and the Internal Number
The final discipline of multi-site OTIF tracking is reconciliation with the customer scorecard. Every major customer in a regulated or service-sensitive industry maintains a vendor scorecard, and the number on that scorecard is what determines contract terms, expediting fees, and the willingness to award new business. Internal OTIF that systematically diverges from customer OTIF is a sign that the internal definition has been bent to make the operation look better.
The reconciliation work is rarely glamorous. It involves agreeing with each customer on the date that counts, the quantity that counts, and the boundary of what counts as a single order. It involves auditing internal records against customer receipt confirmations and tracking down the differences case by case. It involves an immutable record of dispatches and receipts that both sides can point to without ambiguity. The payoff is that the internal number and the customer number converge, and the operations team can act on the metric with confidence that the customer will see the same result.
Manufacturers who do this work earn a different kind of conversation with their customers. Quarterly business reviews stop being defensive exchanges about whose data is right. They become joint diagnostic sessions about where the misses came from and what to do about them. The OTIF customer scorecard becomes a shared artifact rather than a contested one.
The Multi-Site Maturity Curve
Manufacturers move through a predictable maturity curve on multi-site OTIF. Early on, the metric is a single combined number with vague definitions, and it is reported monthly with a wide gap between internal and customer views. The first improvement is decomposition into separate on-time and in-full signals, which immediately exposes which problem dominates. The second improvement is per-order rather than per-shipment counting, which closes most of the gap with customer scorecards. The third improvement is ATP discipline at the commit, which prevents the metric from being damaged by avoidable over-commitments. The fourth is alerting on at-risk orders, which shortens the recovery cycle and lifts the metric without changing the underlying capacity.
Each step requires more discipline than the last, and each one moves the operation closer to a state where the OTIF number means something that customers will recognize. Multi-site networks that complete the curve discover that they have a fundamentally different relationship with their customer base. Networks that stop at decomposition leave most of the value on the table. The metric is not the work. The metric is the score that tells the work whether it is succeeding. For more on the planning side that drives the in-full half of the score, see the explainer on mrp planning horizons.
FalOrb helps multi-site manufacturers track OTIF against an immutable ledger of stock, transfer, and dispatch events with available-to-promise discipline and at-risk order alerting. Visit falorb.com, book a 30-minute walkthrough, or email us at [email protected] to see how it applies to your operation.