← Back to Field Notes

February 19, 2025 · The Bloomfield Team

5 Production Scheduling Mistakes That Kill On-Time Delivery

On-time delivery below 90% almost always traces back to scheduling. The floor is not the problem. The schedule is. Operators run the jobs they are given in the sequence they receive them. When that sequence is wrong, when capacity is overloaded, when material is not available at the time the job is supposed to start, late deliveries are the inevitable result of upstream decisions.

Here are the five scheduling mistakes that cause the most damage. Each one is fixable with better data, better process, or both.

Mistake 1: Scheduling to Theoretical Capacity

A CNC mill available for one shift has 8 hours per day, 5 days per week, 40 hours. That is theoretical capacity. Nobody gets 40 productive machining hours out of a machine running one shift. Setup time takes 10 to 20 percent. Planned maintenance takes 5 percent. Unplanned downtime averages 5 to 10 percent. First-article inspection holds, tool changes, program loading, and operator breaks consume another chunk.

Realistic productive capacity for a single-shift machine in a high-mix job shop is 24 to 30 hours per week. Scheduling 38 hours of work against a machine that can realistically produce 28 creates a 10-hour weekly deficit. That deficit shows up as late jobs by Wednesday of every week. The fix is straightforward: schedule against demonstrated capacity, not theoretical capacity, and track the gap monthly.

Mistake 2: Treating Every Job as Equal Priority

First-in-first-out scheduling is fair and simple. It is also wrong for most job shops. A $5,000 commercial job due Friday and a $50,000 aerospace job due Friday are not equal. A repeat customer who represents 15% of annual revenue and a one-time buyer who found you on Google are not equal. A job with outside processing that needs to ship to the heat treater by Wednesday and a job that runs entirely in-house have different scheduling constraints.

Priority scheduling requires three inputs: delivery urgency, customer value, and process constraints. The shops that maintain OTD above 95% typically run a daily dispatch meeting where the production manager reviews the day's priorities against these three factors and adjusts the sequence as needed. Ten minutes per morning. The return is measured in percentage points of OTD.

Mistake 3: Invisible Queue Time

A job has four operations: mill, turn, grind, inspect. Each operation takes two hours. The estimated total production time is eight hours. The actual throughput time is four days. The gap between eight hours of production and four days of calendar time is queue time, the hours a job spends sitting in a bin next to a machine waiting for its turn.

Queue time is invisible in most scheduling systems because it does not appear as a line item. The ERP shows operation times. The schedule shows start and end dates. Nobody tracks the 6 to 12 hours a job spends waiting between each operation. In a typical job shop, queue time represents 60 to 80 percent of total throughput time. Reducing it by even 20% compresses lead times by days. For a deeper look at why late shipments actually start in queue, that piece covers the mechanics.

Mistake 4: No Visibility into Work-in-Process

Where is job 3847 right now? If the production manager has to walk to the floor and look for it, the scheduling system has failed. Scheduling without real-time WIP visibility is like driving without a speedometer. You know your destination and your starting point. You have no idea how fast you are getting there or whether you are on track.

Shops without WIP tracking cannot identify bottlenecks forming in real time. They discover them when a job is already late. By then, the only option is expediting, which means disrupting other jobs on the schedule and creating a cascade of new late deliveries. Visibility solves this by surfacing the problem when there is still time to adjust. The tools for production visibility connect to existing data in the ERP and give the scheduler what they need to make decisions before problems compound.

Mistake 5: Not Accounting for Setup Sequence

Running job A followed by job B takes 45 minutes of setup. Running job B followed by job A takes 15 minutes because the tooling and fixturing overlap. The sequence in which jobs run on a specific machine affects total productive time by 10 to 25 percent depending on how much setup commonality exists between consecutive jobs.

Optimizing setup sequence requires knowing what tooling each job needs and grouping jobs with similar requirements together. This is trivial to describe and difficult to execute manually when a machine has 15 jobs in its queue. The production scheduler would need to compare tooling requirements across every possible sequence, which is a combinatorial problem that grows exponentially with queue depth. This is where computational tools, whether a simple scheduling algorithm or a more sophisticated AI-based system, create real value by evaluating thousands of possible sequences and finding the one that minimizes total setup time.

The Compounding Effect

These five mistakes do not operate in isolation. A schedule loaded to theoretical capacity with FIFO sequencing, no WIP visibility, unmanaged queue time, and random setup sequences will produce an OTD rate in the low 70s. Every late job that gets expedited disrupts the schedule further, creating more late jobs. The system enters a death spiral where the production manager spends their entire day firefighting instead of planning.

Fix these five problems in order of impact. Start with capacity loading. Move to priority-based dispatch. Add WIP tracking. Begin measuring and reducing queue time. Optimize setup sequences. Each fix builds on the previous one. Shops that address all five systematically see OTD improve from the low 80s to the mid-90s within two to three quarters.