The plant is operating near capacity. Machine reports look respectable. Yet backlogs keep growing, priorities change every day and Sales has to explain missed delivery dates. Simply adding shifts can make matters worse: releasing more work initially creates more work in process (WIP). Customers receive goods sooner only when the right orders move faster through the entire process to shipment.

The management question is: Where is saleable output created, and where is the organization merely keeping people and equipment busy?

Machine utilization is not plant throughput

Consider three successive operations. Under current conditions, the first can deliver 150 good units a day, the second 100 and the third 130. If order release continuously feeds 130 units into that sequence, the queue before operation two will grow. Raising utilization at the first operation will not raise system output. It occupies space, ties up capital and makes priorities harder to manage.

This is deliberately simplified. Product mix, scrap, changeovers and available hours change capacity. For that reason, determine the effective bottleneck for each relevant product family and period rather than relying on a machine’s theoretical master-data capacity.

Little’s Law describes the relationship between WIP, throughput and lead time. In a stable system, average WIP = average throughput × average lead time. If an area carries an average of 600 units in process and completes 100 good units per working day, average lead time is six working days. This is a relationship, not a promise for any individual order. A growing backlog or changing mix requires a more detailed analysis. See the MIT discussion of Little’s Law in manufacturing.

Five reasons why full halls coexist with late orders

1. Orders are released too early. Every department receives work, but queues build at scarce resources. The next urgent order must work its way through material already started.
2. The bottleneck loses productive time. Missing material, quality approvals, unplanned failures and avoidable changeovers directly reduce possible output at the constraint.
3. Local metrics create the wrong incentives. A machine runs a large batch to improve OEE or unit cost while the customer’s order remains behind schedule.
4. Mix and capacity are planned separately. A nominally free hour is of little use when the required resource, skill or material combination is unavailable.
5. Priorities change without a clear decision. Expedited jobs, last-minute replanning and exceptions displace the agreed sequence. Effort rises and reliability falls.

No single metric can identify the dominant cause. Review order flow, material availability, shift calendars and actual loss reasons together.

What I would investigate in the first 30 days

First, make the flow visible for two or three relevant product families. Record release, processing, waiting, rework and shipping events along the actual route. Where is material waiting and for how long?

Second, verify the constraint. A pile in front of a machine is a signal, not proof. The machine may be starved of the right material, blocked by downstream congestion or constrained only for a certain mix or shift.

Third, adapt release and priority rules to effective bottleneck capacity. A deliberately sized buffer protects the constraint from running empty without filling the plant with excess WIP. Teams need clear rules for urgent orders and daily deviations.

Put these measures on the table together

• Good output at the constraint: What did the scarcest resource actually deliver for customer flow?
• Lost time by cause: How many minutes were lost to failures, setup, materials, quality or blocking?
• WIP and age before the constraint: Is the buffer protected, or is a queue growing without benefit?
• Lead time and on-time delivery: Does the improvement reach the customer?
• Schedule changes and backlog: Is the agreed sequence stable enough to produce reliably?

OEE remains useful for analysing technical losses. As the plant’s sole measure of success it is insufficient: a non-critical machine can improve its OEE while pushing more material into a queue. Production KPIs should be interpreted as a connected system, an approach reflected in NIST research on the relationships between manufacturing KPIs.

Turn diagnosis into a management decision

If release is the problem, start fewer orders and finish more. If bottleneck capacity is missing, changeover sequence, maintenance windows, staffing or targeted additional hours may help. If the constraint shifts with product mix, demand, capacity and priority decisions belong in S&OP/IBP; daily S&OE keeps implementation on course.

By day 30, establish a credible baseline. By day 60, pilot release rules and constraint management in a defined area. By day 90, use good output, lead time, WIP and on-time delivery to decide what to scale. These are decision gates, not a blanket performance guarantee.

The standard is not how busy every machine looks. It is how reliably the plant completes the right orders. The companion insight Production Bottlenecks: Which KPIs Actually Drive Decisions? goes deeper into the scarce resource, from good units and lost minutes to customer delivery and contribution.

For an initial self-assessment, the Production & Operations Health Check offers a free Basic Check without sign-up and optional monitoring by email. The related English checks cover end-to-end supply chain maturity, demand and capacity decisions and inventory and working capital. Operational and cash impacts must be validated against plant data.