Two artifacts sit on the same desk, 18 inches apart. On the left, this week's production schedule, printed Monday at 6 a.m., every job assigned to a work center and a day. On the right, last week's shipping log. Read them line by line and they describe two different plants. Jobs that shipped a week before their scheduled date. Jobs scheduled for Wednesday that went out nine days later. Put them together and the distance between the plan and the week becomes hard to miss.
A third document actually ran the floor: the marked-up copy, the same printout on a clipboard or in a plastic sleeve, with three jobs circled, one crossed out, an arrow moving something off the mill, and a phone number in the margin. It gets marked up before the first setup Monday morning. By Thursday nobody's looking at it, because everybody in the building already knows the real order.
The distance between the schedule and the shipping log is an evidence problem. Nobody has measured schedule adherence, so nobody in the building can say how large the gap between the plan and the output really is. They only know it closes every week without anyone asking.
A plant that misses its printed schedule doesn't just run late. Somebody pays to bring it back, and the instruments are ordinary. Overtime Thursday and Saturday. Expedite freight on the jobs that can't slip. An outsourced secondary operation at a price nobody would have quoted. The owner's own weekend, spent making the calls the schedule was supposed to make unnecessary.
Jill Jusko, writing in IndustryWeek, describes overtime as "the fix, the expensive fix, to a multitude of manufacturing miscues that deserve better and more permanent solutions," and names poor production planning, which can drive bottlenecks as one of the things it covers. That tells us what overtime can hide, not how often it hides it.
What makes it hard to see from inside is that the financing never asks permission. A capital request gets a form, a discussion, and a signature. An expedite freight bill gets coded to freight. Overtime gets coded to labor. Both land in the same buckets they'd land in if the week had gone perfectly, at slightly larger numbers, and nobody has to explain either one. The spending is real and recurring, and it's invisible in the only way that counts: nothing in the reporting ever forces the question of what it bought.
Start with the staffing arithmetic. Permanent headcount costs money in quiet weeks; overtime costs money in busy ones. When demand genuinely swings, premium hours can be cheaper than carrying the peak all year. An owner who makes that choice deliberately is doing arithmetic, not hiding from anything, but the defense has to show that base hours were already being used and that the premium hours bought flexibility the plant actually needed. The consultant who walks the floor and reads the overtime line as a defect has skipped the only question worth asking.
Which kind do you have? You can't answer that from the overtime line, because both kinds look identical there.
Two measures, both built from data the plant already has.
First, schedule adherence per work center over a fixed window. Not plant-wide. Per work center: jobs completed in the period they were scheduled, at the center they were scheduled on. Pick the window, pick the centers, hold the definition still.
Second, reconcile routed standard hours against actual hours booked by operation, then look for the difference in a short list of places. Setup time the routing understates. The second setup after a scrap event. Secondary operations that happen on every job and exist on no routing. Releases pulled in because a customer asked and nobody re-planned around it. Rework re-runs. Hours where a machine was staffed and idle, waiting on tooling or material.
This is a data question before it is a software question, and an owner who now wants to know what the system of record is actually supposed to do for him is better served by our piece on shop-floor technology than by a vendor's demo. The work is weeks of honest extraction, and the extraction isn't the hard part. The hard part is organizational. The first number may look bad, and every instinct in the building, including yours, will want to redefine the measure until it looks better. Hold the definition. Record every exclusion with its reason. Keep the routing revision history, so a corrected standard can be told apart from a convenient one.
Give the adherence number to the plant leader or scheduler and the reconciliation to the controller. Your job is to decide what the result changes. If the spreadsheet comes back to you every week, you have simply built another version of the same problem.
Run the reconciliation and you learn something about capacity you probably already suspected. The floor hits its dates because a particular person knows which jobs can slide, which customer will take a phone call instead of a shipment, which setup can be split, and which machine to steal an hour from. That's judgment. It's usually good, and it produces results.
It also means usable capacity is currently an attribute of two people rather than of the plant. Same species of problem as the machinist whose retirement takes the process with him, except the scheduler's version is harder to see, because his output looks like a normal week.
A buyer may find it for the dullest reason: the checklist says to look. Michael L. Vaccarella of Wipfli, writing diligence guidance for buyers of manufacturing companies, tells them to dig behind the reported numbers and asks flatly: "Is your shop floor running at capacity? Or are hidden bottlenecks and snags hindering productivity?" That's a professional services firm instructing its buy-side clients to test this exact thing. When we run that routing check and find a mismatch, we do not attach a predetermined number to it. We start testing every capacity claim the owner has made, including the ones that were true, which is the part that stings.
There is no magic in the routing check. Two exports and a pivot table can make an outsider look unusually perceptive. They simply create the one place where an owner's confidence and the plant's documentation can be laid next to each other, and the owner has never had a reason to lay them there himself.
Not ready and not sellable are different diagnoses, and it's worth being careful about which one you're looking at. A plant with routings that overstate capacity can be entirely sellable. A buyer may bring capital, systems, or a second scheduler you never chose to build. The plant can have plenty of merit and still lack the evidence to prove it.
So the recommendation is narrower than it sounds. Correct the routings the reconciliation proves wrong. Document the measured adherence and keep measuring it. Don't launch a pre-sale improvement campaign, because a half-finished scheduling initiative in the middle of diligence reads worse than an honest number with a trend behind it. And don't fix the things that would only ever be fixed for an audience: the adherence number on a work center you're planning to retire, the standard on a part you run twice a year.
An owner who's moved past diagnosis and toward the question of what a process actually involves will get more out of the founder's guide to selling a manufacturing business than out of another operating article.
The other outcome is that you measure this, correct four routings, and cut zero overtime hours in the first quarter. That's a fine result. You're in a stronger position anyway, because now you can say which hours you chose.
Go back to the clipboard. The first read on that sheet is that it's a symptom, evidence of a system nobody trusts.
The second read is more accurate. That marked-up sheet is the plant's real operating system. It's current, it's correct, it gets updated weekly without anyone being asked, and it works. It's also readable by exactly one organization, which is this one, and only while the person holding the pen keeps showing up.
Keep marking it up. Pens are fine. Just make sure something else in the building would still know the schedule if the pen stopped.