The standard advice on manufacturing metrics is that you should measure what matters, align your KPIs to strategy, and make sure every level of the organization has visibility. It appears in every operations textbook and it is not wrong. It is just useless, because it does not tell you which metrics to pick, and the reason companies end up with a forty-metric dashboard is not that they disagreed with the advice. It is that they followed it forty separate times.
Here is what actually happens when you add a metric. Someone has to produce it, which costs time. Someone has to look at it, which costs attention. And crucially, it gets a line on a scorecard, which means it acquires a claim on the organization’s response. A metric on a scorecard is a standing instruction to react when it moves. Add forty of them and you have issued forty standing instructions to a management team that can genuinely act on maybe six things at once. What you get is not forty-metric management. It is six-metric management chosen arbitrarily each week by whoever is loudest, plus thirty-four metrics generating quiet anxiety.
So the question is not what matters. Almost everything matters a little. The question is which twelve numbers earn a permanent claim on management attention in a manufacturing business, and that is a much harder and much more useful question.
My answer comes from running plants and businesses at very different scales, including four years owning a small industrial tank liner manufacturer that I bought, operated, and sold to its largest vendor, and my current role running a roughly one billion dollar division at JBT Marel. Twelve metrics, four families. The families matter more than the individual picks, because the families are what tell you when your scorecard has a hole in it.
Why Does Adding Metrics Make a Business Slower?
Because measurement is not free and the cost is not the reporting effort. The cost is response capacity.
Every metric on a reviewed scorecard creates an implicit obligation: if this moves badly, someone will ask about it, and someone will have to do something. That obligation is finite and it is consumed in order of visibility rather than in order of importance. A metric that moves dramatically will pull response capacity away from a metric that moved slightly in a direction that actually costs more money.
This produces a specific and very common failure. A business with a large dashboard reacts reliably to volatile, visible, easily-measured things, and reacts slowly or never to stable, structural, hard-to-measure things. Scrap spikes get immediate attention. A product line that has been quietly losing two points of margin a year for four years gets a footnote. The dashboard did not cause the misallocation, but it legitimized it, because every week the team can point at a screen showing that it is monitoring the business.
The second cost is that large scorecards prevent disagreement. With twelve metrics, a leadership team can argue about which one is the binding constraint this quarter, and that argument is enormously productive. With forty, nobody can hold the whole picture, so the argument never happens and the team defaults to whichever function speaks with the most confidence.
A metric on a scorecard is a standing instruction to react when it moves. Issue forty of them to a team that can act on six things, and the other thirty-four just generate anxiety.
What Are the Four Families a Scorecard Needs?
A manufacturing business makes and keeps a promise, converts material into product, absorbs the cost of the times it gets that wrong, and either makes money doing it or does not. Those are the four families, and every genuinely useful manufacturing metric sits in one of them.
Promise covers whether customers can count on you. Flow covers how fast and how reliably material converts. Failure cost covers what going wrong actually costs, as opposed to how often it happens. Money covers whether the first three families produced a return.
The value of thinking in families rather than in metrics is diagnostic. Look at your current scorecard and sort every metric into one of the four. In most mid-market manufacturers the distribution is badly lopsided: heavy on flow, moderate on failure, thin on promise, and almost empty on money below the total P&L line. That shape tells you exactly where the business is blind, and it is a far faster diagnosis than debating individual metric choices.
Which Metrics Tell You If Customers Can Count On You?
On-time in full. Measured against the date you originally promised, not the date you rescheduled to. This distinction is where most on-time reporting quietly becomes fiction. A business that reschedules and then measures against the new date can report excellent performance while its customers experience chronic unreliability, and the customers are the ones who decide whether to reorder. Measure against the original commitment, take the lower number, and manage the real problem.
Quoted order lead time. The elapsed time from order receipt to shipment, tracked as a distribution rather than an average. The average is nearly worthless here because customers do not experience your average, they experience their own order. What you want to know is the spread: what does the slowest tenth look like, and is the spread widening? Widening spread is an early warning of a capacity or scheduling problem that has not yet shown up anywhere else.
Past-due backlog dollars. Not units, not line items, dollars. This converts a service problem into a cash problem and makes it visible to people who were ignoring it when it was expressed in units. It is also the single fastest read on whether the plant is falling behind, because it accumulates before any of the flow metrics deteriorate enough to trigger attention.
Which Metrics Tell You How Fast Material Moves?
Schedule attainment. Did you build what you said you would build, in the sequence you said you would build it? Not did you build the right total volume. Sequence matters because the whole downstream promise depends on it, and a plant that hits volume by substituting easy jobs for hard ones is generating a future service failure while reporting a good week.
Constraint utilization. Utilization of the one resource that actually limits output, and nothing else. Utilization measured across all equipment is worse than useless, because it rewards running non-constraint machines to look busy, which produces work in process that has to be stored, moved, counted, and eventually written off. Find the constraint. Measure it. Ignore utilization everywhere else.
Inventory turns. By category, not in aggregate. Raw, work in process, and finished goods fail for completely different reasons, and a blended turns number hides which one is the problem. Rising work in process with stable finished goods is a flow problem. Rising finished goods with stable work in process is a demand or a mix problem. The aggregate number tells you neither.
Which Metrics Tell You What Failure Costs?
First-pass yield. The percentage that goes through the process correctly the first time, with no rework, no touch-up, no sorting. Most quality reporting measures what shipped correctly, which includes everything that was caught and fixed. That is a measure of your inspection, not your process. First-pass yield measures the process, and the gap between the two numbers is precisely the cost of the inspection you are paying for instead of the capability you do not have.
Cost of poor quality. Scrap plus rework labor plus warranty plus the expedited freight that quality problems generate. Assembled into one dollar figure and reported monthly. It is uncomfortable and it is the most persuasive number in the plant, because it converts a technical conversation that operations owns into a financial conversation that everyone owns.
Unplanned downtime hours. Hours, on the constraint, unplanned only. Keeping it in hours rather than a percentage keeps it legible and keeps it addable to a capacity conversation. Planned maintenance is a decision; unplanned downtime is a failure, and blending them lets the failure hide inside the decision.
Measuring what shipped correctly is a measure of your inspection, not your process. The gap between that and first-pass yield is what your missing capability costs you every month.
Which Metrics Tell You Whether Any of It Made Money?
Gross margin by product line. This is the metric most commonly missing and most commonly worth the most. Total gross margin tells you almost nothing about what to do, because it is an average across lines that are usually performing very differently. The method I use for pulling the truth out of it is the 80/20 Matrix of Profitability, which sorts the business by where profit actually concentrates rather than where revenue does, and the result is regularly uncomfortable for whoever has been defending the long tail.
Price realization. Realized price against target or list, by line and by customer. This is where margin leaks without anyone deciding to let it leak, through discounting authority that was granted years ago and never reviewed, through freight terms, through rebates that accumulate quietly. Price realization makes the leak visible as a single number and it is frequently the fastest margin recovery available in a mid-market manufacturer.
Cash conversion cycle. Days of inventory plus days sales outstanding minus days payable outstanding. One number that captures how long your money is trapped inside the operation. It belongs on the operating scorecard, not just the finance report, because every one of its three components is driven by operating decisions that the plant makes without seeing the cash consequence.
Which Popular Manufacturing Metrics Mislead?
Overall equipment effectiveness is the biggest one, and it deserves its own treatment rather than a paragraph here. The short version is that it aggregates three distinct failure modes into a single number, which makes the number nearly impossible to act on, and it invites optimization of non-constraint equipment where improvement produces no additional output at all.
Labor efficiency and absorption belong in the same category. Both reward production volume irrespective of whether the volume was needed, which means both will look excellent during exactly the period when you are building inventory you cannot sell. Any metric that improves when you overproduce is a metric that will eventually cost you cash.
Cost per unit is the most seductive of the misleading metrics because it sounds like the truth. It falls when volume rises, since fixed cost spreads across more units, which means a plant can improve cost per unit by building product nobody ordered and can appear to deteriorate during a month when it correctly cut production to match demand. Managed hard enough, it will reliably produce inventory. If you keep it, keep it alongside the volume it was calculated at, and never compare two months at different volumes without saying so.
Capacity utilization across the whole plant deserves the same treatment as equipment utilization: it rewards activity at non-constraint resources, where additional output has no value and creates work in process that has to be handled and carried. The version worth tracking is utilization of the constraint alone, which is one of the twelve.
Revenue per employee is a board metric that has migrated onto operating scorecards where it does not belong. It moves with price, mix, and the make-versus-buy decision far more than with anything a plant controls, which means it will look excellent in the year you outsource a component and terrible in the year you insource one, independent of whether either decision was right.
Average anything deserves suspicion. Average lead time, average cycle time, average yield across products. Averages are built to hide variation and variation is the thing you are usually trying to manage. Where a metric is reported as an average, ask for the distribution once and watch what turns up in the tail.
How Many Metrics Should Actually Be on the Scorecard?
Twelve at the enterprise or business unit level. Fewer further down, not more.
The instinct at the plant or department level is to cascade all twelve and then add local detail, which produces a twenty-metric departmental scorecard and defeats the purpose. The right cascade is narrower: each level owns the three or four metrics it can genuinely move, plus visibility into the level above so it understands what its contribution feeds. A department that can act on first-pass yield and schedule attainment should be measured on first-pass yield and schedule attainment, and should not be held to a cash conversion cycle it has no lever on.
Twelve is not magic. It is roughly the number a leadership team can carry in its head, argue about, and act on within a single operating cycle, which is the actual constraint. If your team can hold fifteen, hold fifteen. Nobody holds forty.
How Do You Replace a Forty-Metric Dashboard Without a Fight?
You do not announce that thirty metrics are being deleted. You announce that twelve are being elevated, and you let the rest keep existing in the reporting layer where they cost almost nothing.
That distinction is the whole trick, and it is not a political dodge. The problem was never that the other metrics exist. Data is cheap. The problem is that they occupy scorecard status, which is what creates the claim on response capacity. Move them to a reference report that anyone can pull, keep producing them, and stop reviewing them in the meeting. The finance analyst who built a metric three years ago keeps their work. The leadership team gets its attention back. Nobody has to defend a deletion.
Run the transition for one full quarter with both structures visible, and at the end of the quarter ask a single question: which of the thirty reference metrics did anyone actually pull, and for what decision? The answer is usually three or four, and those three or four are candidates to replace something in your twelve. The rest have now demonstrated, with evidence rather than argument, that they were never driving anything.
Expect two objections. The first is that something important will be missed, which is answered by threshold alerts: any reference metric can carry a trigger that escalates it into the meeting when it crosses a defined line, so nothing goes unwatched, it just stops being discussed by default. The second objection is that a function will lose visibility with senior leadership, and that one is real. The answer is to give each function a standing slot to bring a case built on its own reference data, rather than a permanent scorecard line it has to justify every month.
Who Should Own Each Metric?
One named person per metric, and that person must have authority over at least one lever that moves it. An owner without a lever is a reporter, and putting a reporter’s name on a metric guarantees that bad movement produces an explanation rather than an action.
The test is worth running deliberately. For each of the twelve, name the owner, then name the specific decision that owner can make to move the number. If you cannot name the decision, the ownership is misassigned, and the usual cause is that the metric has been assigned to whoever produces the report rather than to whoever controls the process.
Cross-functional metrics are where this breaks most often. On-time in full is driven by sales promising, planning scheduling, and operations building. Assign it to one of the three and the other two will point at that one. The resolution is not shared ownership, which is no ownership. It is to assign the metric to the person accountable for the outcome, and to make the review structure force the other functions into the room. That is a job for the operating cadence, not for the scorecard.
How Do You Set a Target Without an Industry Benchmark?
You use your own distribution, which is better than an industry benchmark anyway.
Take twelve months of your own history for the metric and look at the best month you actually achieved. You did it once, under real conditions, with your real equipment and your real people. That is a demonstrated capability, and a target set slightly inside your own best demonstrated month is both credible and defensible in a way that a borrowed industry figure never is. Nobody can argue it is impossible. You already did it.
The second reference point is your own variation. If the spread between your best and worst month is wide, the first target is not a better average, it is a narrower spread. Consistency usually has to be won before level does, and teams that chase the average while ignoring the variance tend to produce a heroic month followed by a collapse.
Industry benchmarks have a real use, but it is strategic rather than operational. They tell you whether the business is structurally competitive. They do not tell a plant manager what to do on Tuesday, and treating them as operating targets produces goals that the team privately regards as fictional, which is worse than having no target at all. The question of how often each of these numbers should be reviewed is a separate discipline, and I have written about matching review frequency to the metric elsewhere.
One limitation worth stating plainly: twelve metrics is a scorecard, and a scorecard tells you where to look, not what is wrong. Every one of these numbers will at some point move for a reason that has nothing to do with the process it is supposed to describe, usually mix. Any scorecard used as a substitute for going to the floor and asking will eventually produce a confident, well-measured, entirely incorrect conclusion.
About the Stagnation Assassin
Todd Hagopian is a Fortune 500 transformation executive who has generated $3B in shareholder value across Berkshire Hathaway, Illinois Tool Works, Whirlpool, and JBT Marel, where he serves as Global VP of Product Strategy and Marketing. Known as The Stagnation Assassin, he is the author of two published books: The Unfair Advantage: Weaponizing the Hypomanic Toolbox and Stagnation Assassin: The Anti-Consultant Manifesto. His blog is published in 15+ languages and read by operators worldwide. Bring him to your stage via his speaking page or connect with him on LinkedIn.
15-minute 80/20 Portfolio Performance Audit. Send me gross margin by product line and I will show you where profit is actually concentrated, which lines are consuming capacity without earning it, and what the scorecard is hiding. Free, fifteen minutes, no pitch. Request the audit here.
See also: How Often to Review the Numbers