- What a plant turnaround actually is, and what it is not
- Before you change anything: the first two weeks
- Stabilize cash before you improve anything
- Get visibility you can trust
- Fix the decision layer
- Attack margin in order of speed
- Deal with the portfolio drag
- Build the operating cadence that holds the gains
- The capability problem you have to solve by month three
- What month twelve should look like
- Frequently asked questions
The first manufacturing business I owned outright made industrial tank liners. Not a division, not a P&L I was accountable for, but a company I bought with my own money, ran, and eventually sold to our largest vendor. The reports I inherited at closing were tidy. The floor was not. Those two things describe most underperforming plants I have walked into since, and the gap between them is where the entire turnaround lives.
My name is Todd Hagopian. I have generated $3B+ in shareholder value across Berkshire Hathaway, Illinois Tool Works, Whirlpool, and JBT Marel, and most of that came from businesses that were not broken in any dramatic way. They were slow. They were opaque. They were losing a point of margin a year to things nobody had measured in a decade. This article is the full sequence I run, from the first walk of the floor through the twelve-month mark, in the order that works.
Short answer: A plant turnaround is a sequencing problem, not a knowledge problem. Stabilize cash, then restore measurement you can trust, then fix the decision layer, then recover margin from fastest lever to slowest, then install a cadence that prevents the slide back. Skipping to margin work before you can trust the numbers is the single most common way these fail, and it is why so many plants get fixed twice.
What a plant turnaround actually is, and what it is not
A turnaround restores a plant’s ability to convert demand into profitable output at a rate the business can fund. That is the whole definition. Everything else people call a turnaround is a different activity wearing the word.
Three things it gets confused with
A restructuring changes the capital structure. It moves debt, renegotiates covenants, sells assets, and buys time. It can be necessary and it can run alongside a turnaround, but it does not change how the plant runs on Tuesday. If you fix the balance sheet and leave the operating system alone, you have bought eighteen months and spent them.
A cost reduction program is a lever inside a turnaround, not the turnaround itself. Cost programs have a defined end and a defined number. Turnarounds have a defined end state. The difference matters because a cost program applied to a plant with a demand problem shrinks the business permanently, and I have watched that happen to operations that were three quarters away from being fine.
A continuous improvement rollout is what you do after. Kaizen, standard work, and the rest of the toolkit assume a stable baseline and a management layer with the bandwidth to sustain them. Underperforming plants have neither on day one. Deploying them first is how you get a wall of laminated A3s and a flat margin line.
The order that works
The sequence below is not a maturity model and it is not a list of workstreams to run in parallel. Each stage exists because the one after it is impossible without it. You cannot recover margin from a line whose output you cannot measure. You cannot measure a line whose supervisor is spending his day chasing a decision that is stuck two levels up. And you cannot ask anyone to do any of it while the business is three weeks from a cash event.
The method I use to sequence this work is the HOT System. It orders the intervention by dependency rather than by urgency, which is why it survives contact with a plant where everything is urgent.
If you want a useful counterweight to my sequencing, the Harvard Business Review piece on how to turn around nearly anything makes the case that the diagnosis phase should be longer than most operators allow. I agree with the principle and disagree with the calendar. Two weeks of counting is enough if you count the right things, and every week past that is a week the organization spends watching you decide whether you are serious.
Before you change anything: the first two weeks
The first two weeks buy you the right to act, and you spend them establishing what is true rather than what is reported. This is not a listening tour. A listening tour produces a slide deck of themes. What you need is a small number of counted facts that contradict the reporting, because those contradictions tell you where the operating system has quietly stopped working.
Count things yourself
Pick three things and count them with your own eyes, at least twice, on different shifts. Units off the constraint line in an hour. Machines actually running during a randomly chosen fifteen-minute window. Pieces sitting in the WIP buffer before final assembly. Then compare each one to what the system says.
The number you get is less important than the size of the gap. A plant where your count matches the report has a demand or a pricing problem and you can move quickly to the commercial side. A plant where your count is materially different from the report has a measurement problem, and every improvement initiative running there is pointed at a target that does not exist.
Establish four numbers before you speak publicly
Cash on hand and the weekly burn, so you know how much time you have. Output per labor hour on the constraint, so you know the plant’s true capacity. Gross margin by product family rather than in aggregate, because the aggregate hides both the winners and the bleeders. And on-time delivery measured from the customer’s requested date, not from the date your own system reset it to after the first miss.
That last one causes more arguments than the other three combined, and it is usually where you find that the delivery performance the sales team has been apologizing for is actually worse than the number they were apologizing about.
What not to do in the first two weeks
Do not reorganize. Do not announce a strategy. Do not cancel a project you do not yet understand, because half the projects in a struggling plant exist to work around a problem nobody has told you about, and killing the workaround before you fix the problem makes things worse in about ten days.
Do say clearly and early what you are doing and when people will hear from you. Ambiguity in week one gets filled with the worst available rumor, and you will spend month two undoing it. If you want the seven underlying causes to look for while you are counting, I laid them out in detail in why manufacturing plants underperform.
The one conversation to have before week two ends
Everything above happens inside the building, and if you stop there you will finish your diagnosis with an operations answer to a question that may not be an operations question. So before the two weeks close, sit down with whoever owns the commercial side and ask three things.
What did we quote in the last four quarters and what share of it did we win. Where the losses were, ask what the customer told us the reason was, and press past the reflexive answer of price, because in industrial businesses the stated reason is price and the actual reason is frequently lead time or a past delivery failure that nobody logged. Then ask which customers have quietly reduced their share of wallet without ever leaving, since that decline never appears as a lost order and never triggers a review.
What you are testing is whether the plant’s problem originates upstream of the plant. An operation that is winning the work it quotes and failing to deliver it has an execution problem, and everything in this article applies. An operation that has stopped winning has a position problem, and applying an execution recovery to it produces a very efficient plant with nothing to make. That distinction is worth two hours in week two, and it will save you two quarters later.
Stabilize cash before you improve anything
Cash comes first because cash buys the calendar, and the calendar is the only resource a turnaround genuinely cannot manufacture. Every other constraint can be worked around by a competent operator. Running out of time cannot.
Why this is not the same as cost cutting
Cash stabilization is largely about timing and working capital rather than about spending less. Most of what you release in the first six weeks is money the business already owns and has parked somewhere unproductive. That distinction matters politically as well as financially, because releasing trapped cash does not require you to take anything away from anyone, and it lets you demonstrate momentum before you have asked the organization for a single sacrifice.
The five places cash is usually hiding
Inventory that exists to cover for a process nobody trusts. Safety stock is a tax the plant pays for its own unreliability, and it is almost always sized off a reliability number from years ago. Receivables that have drifted past terms because nobody owns the collection conversation and the sales team considers it a relationship risk. Premium freight and expedite fees, which are the cash cost of a planning system that has stopped working and which almost never appear as a line anyone reviews. Capital projects in flight that were approved under a different set of assumptions and that will not deliver inside your window. And payables terms that were never renegotiated after volume changed.
Hypothetical: if a plant carries 90 days of inventory and the process genuinely requires 60, releasing those 30 days at a 20% cost of goods run rate frees roughly one twelfth of annual COGS in one-time cash. On a $60M COGS base that is about $5M. Those figures are illustrative and are meant to show the shape of the arithmetic, not a benchmark for your business.
The rule for this stage
Take every action that releases cash without permanently reducing capability, and take none that do. Selling a machine you will need in eight months to make this quarter’s number is not stabilization. It is a second turnaround scheduled for next year, and the person who inherits it will be able to name exactly who created it.
Get visibility you can trust
You cannot improve what you measure once a month, because a monthly number tells you what happened rather than what is happening. This stage is where most turnarounds are actually won, and it is the one operators skip because it feels like administration rather than action.
Measure at the frequency of the decision
The correct measurement frequency is the frequency at which someone can do something about it. A supervisor decides what to run and who to move within the hour, so the supervisor’s number is hourly. A plant manager decides on overtime, sequencing, and escalation within the day, so that number is daily. A general manager decides on mix, pricing, and capital within the month.
Almost every underperforming plant I have seen measures everything at the general manager’s frequency and then wonders why the floor cannot respond. A shift that ends 400 units short does not know it until the following week, and by then the four separate hours where it went wrong are unrecoverable memory. Move the constraint line to an hourly count posted where the line can see it, and you will get an improvement in the first fortnight from nothing but attention.
One number per line, visible to the people who move it
Resist the dashboard. A dashboard is a device for making a management team feel informed, and its natural failure mode is comprehensiveness. Each production area gets one number that it owns, posted physically, updated by the people doing the work. If the number is being typed into a spreadsheet by someone in an office, it is a report, not a measurement, and it will drift.
Kill the reports nobody uses
Ask for the distribution list and the last time each recurring report changed a decision. In every plant I have run this exercise in, a meaningful share of the reporting burden turned out to be serving a request from a person who had left the business. That labor is your first source of free capacity for the work that follows.
Fix the decision layer
Decision latency is the root cause that hides behind every other root cause. Ask a supervisor what is preventing him from hitting his number and he will name a machine or a material. Ask him what he did about it and you will find a request that has been sitting in someone’s inbox for eleven days.
Find the queue
For one week, log every decision that leaves the floor: what was asked, who it went to, and when it came back. You are looking for the median return time and for the two or three people who appear as the destination most often. The queue is almost never distributed evenly. It is one or two well-intentioned people who have accumulated approval rights over a decade and who are now the plant’s real constraint.
Push the threshold down and put a clock on it
Raise the spending and scheduling thresholds at the supervisor level to a number that would make the CFO mildly uncomfortable, and pair it with a rule that any decision escalated above the floor gets an answer inside 24 hours, including the answer no. A fast no is worth more than a slow yes, because a fast no lets the supervisor go find the second option while the shift is still running.
The standing decision meeting
Thirty minutes, same time daily, standing, with the people who can commit resources in the room. Not a status update. The only agenda item is decisions that are blocked, and the meeting ends when they are unblocked. Within a month, the meeting gets shorter, which is the signal that the layer is working rather than the signal that people have stopped bringing things.
Attack margin in order of speed
Once the numbers are trustworthy and decisions move, you go after margin in the order of how fast each lever pays, not in the order of how large each lever looks. Speed matters here for a reason that has nothing to do with impatience: early wins fund the credibility you need for the slow levers, and the slow levers are where the structural money is.
Levers that move inside 30 days
Discount drift, which is the accumulated concession authority nobody has audited. Premium freight, which stops when someone is required to sign for each instance. Expedite fees charged by suppliers because your planning horizon collapsed. Scrap and rework on the two worst part numbers rather than across the board. Overtime that is scheduled rather than earned.
Levers that move inside 90 days
Price, on the segments where you have a genuine position. Payment terms. Changeover time on the constraint, which converts directly into either capacity or labor cost depending on which one you are short of. Mix, meaning the deliberate steering of quoting and lead time toward the product families you already know carry margin. Labor scheduling matched to actual demand shape rather than to a staffing pattern set when the order book looked different.
Why price belongs at 90 days and not at 30
Price is the fastest arithmetic lever in any manufacturing business and the slowest one to execute safely, and operators under pressure consistently confuse those two facts. The arithmetic is trivial: a point of price falls entirely to gross margin, while a point of cost has to be found, verified, and defended. The execution is not trivial at all, because a price move made before you can measure what it does is a move you cannot evaluate or reverse.
What you need in place first is margin visibility at the product family level and a delivery performance number you believe. Without the first, you will raise price on the wrong segment and lose the volume you were making money on. Without the second, you will raise price into a service failure, which is the one combination that converts a price increase into a customer loss rather than a negotiation.
Both of those arrive during the visibility stage, which is why the sequence puts price at 90 days. Get there and the move is straightforward: segment by where you have a genuine position, move price where you do, hold where you do not, and instrument the win rate weekly so you can see the response inside a month instead of arguing about it inside a year.
Levers that take a year
Footprint. Make versus buy. Product line rationalization. Automation with a real payback. These are the ones that change the plant’s cost structure permanently, and they are also the ones that require the board’s patience and the organization’s trust, which is precisely why they come after the first two groups rather than before.
Before you pull any of them, be certain which problem you are solving. Cost levers applied to a demand problem shrink a business permanently, and the diagnostic for telling them apart is in cost problem or demand problem.
Deal with the portfolio drag
Portfolio drag is the accumulated cost of products the plant makes out of habit, and it is invisible in the P&L because complexity cost hides inside overhead and gets spread evenly across everything. The result is a standard cost system that quietly taxes your best products to subsidize your worst.
Every product family carries a set of costs that standard costing does not attach to it: tooling maintenance, changeover time, the inventory it forces, the forecast error it generates, the documentation it requires, and the share of engineering and customer service attention it consumes. Attach even a rough version of those costs and the profitability distribution stops looking like a gentle slope and starts looking like a cliff.
Complexity cost hides inside overhead and gets spread evenly across everything, which means a standard cost system quietly taxes your best products to subsidize your worst.
In a turnaround you are not running a full rationalization. You are looking for the small number of product families whose true cost to serve exceeds their price, and you are stopping the bleeding on those while the larger portfolio work waits for a calmer quarter. Rough and fast beats precise and never, and a three-week allocation that gets acted on is worth more than a six-month costing project that arrives after you have already left.
Build the operating cadence that holds the gains
The cadence is what converts a turnaround into a permanent change in how the business runs, and it is the stage that separates plants that stay fixed from plants that get fixed twice. A cadence is not a meeting schedule. It is a set of standing questions asked at a fixed frequency by a named person, with the authority to act attached.
Daily
What did we make against what we said we would make, and what is blocked. Thirty minutes, floor level, standing, closed with commitments rather than actions.
Weekly
Where did margin land against plan by product family, what did we spend on premium freight and expedites, and what decisions did not clear within 24 hours. That last question is the health check on the decision layer, and dropping it is how the queue quietly rebuilds.
Monthly
Mix, pricing, capital, and the leading indicators. The monthly review is the only one that should include a document, and the document should be short enough that everyone has read it before the meeting begins.
Quarterly
The honest question: is the trajectory still real, or are we harvesting? A plant can post four good quarters by deferring maintenance and running the constraint hot, and the tell is that output rises while unplanned downtime and scrap rise with it. Ask for those two numbers alongside the good news, every quarter, permanently.
The capability problem you have to solve by month three
By month three the sequence stops depending on you and starts depending on your supervisors and middle managers, and this is where more turnarounds stall than anywhere else. The plan is sound, the numbers are visible, the decisions move, and execution still does not happen, because the layer between strategy and the floor has spent years being told what to do rather than being asked to decide.
The diagnosis is usually simpler than it looks. Middle managers in stagnant plants are not resistant. They are uncalibrated. They have never been given a number they own end to end, never had the authority to move resources against it, and never seen a peer rewarded for making a call that turned out to be wrong but reasonable. Give them all three within the same quarter and a meaningful share of them turn out to be very good at this.
The rest need coaching, and a small number need a different role. Make those judgments deliberately, make them by month four, and make them on evidence rather than on energy in meetings. The most impressive presenter in the room is not reliably the person whose area is improving, and confusing the two is an expensive error that takes a year to surface.
What month twelve should look like
The test at twelve months is not the margin number. It is whether the plant would keep improving if you left on Friday, and there are four observable signals that answer it.
The daily number is produced by the floor without anyone in an office assembling it. Decisions that leave the floor come back inside a day without you chasing them. The weekly margin review is run by someone other than you and includes bad news that you had not already heard. And the plant has caught and corrected at least one stall on its own, without escalation, which is the only genuine proof that the operating system rather than your presence is doing the work.
If those four are true, the turnaround is finished even if the financial recovery is not, because the mechanism that produces the recovery is now installed. If the margin is up and those four are false, you have a result rather than a turnaround, and it will unwind on a schedule set by your successor’s start date.
Frequently asked questions
How long does a manufacturing plant turnaround take?
Cash stabilization takes weeks, trustworthy measurement takes one to three months, and the structural margin work takes most of a year. The twelve-month mark is a reasonable point to judge whether the operating system has changed. Judging it at 90 days measures your energy, not the plant.
What should you fix first?
Cash, then measurement, then the decision layer, then margin. The order is a dependency chain rather than a preference. Margin work performed on numbers you cannot trust produces improvements that do not appear in the P&L, which costs you the credibility you need for the slower levers.
Do you have to cut costs to turn a plant around?
No, and cutting first is the most common way to make an underperforming plant permanently smaller. Establish whether the shortfall is a cost problem or a demand problem before you touch the cost base, because the two require opposite responses and the wrong one is not reversible.
How do you know if the turnaround is working?
Watch leading indicators rather than the margin line. Output per labor hour on the constraint, the median time for an escalated decision to return, premium freight spend, and unplanned downtime will all move before the P&L does. If output improves while downtime and scrap also climb, you are harvesting rather than improving.
Should you hire a turnaround consultant?
Bring in outside help when you need capacity or a specific technical capability you do not have, and keep the sequencing decisions inside the business. The failure mode is not the consultant’s competence. It is that the operating cadence never transfers, so the improvement leaves when the engagement ends.
So the question is how to turn around an underperforming manufacturing plant, and the answer is that you do it in dependency order: cash, then visibility, then decisions, then margin, then cadence. Every plant I have seen fail at this knew what to fix and got the order wrong.
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 VP of Global Product Strategy. 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.
Running this sequence right now? The stage that decides the outcome is the decision layer, and most operators cannot see their own queue from inside it. Book a Decision Velocity Audit and we will measure how long a decision actually takes to clear your plant, and where it stops.