- The job you were actually given
- Why the plant manager’s playbook fails at this altitude
- What should you measure first?
- Which sites are actually yours to fix?
- How do you tell a site problem from a portfolio problem?
- From diagnosis to action
- Days 1 to 30: the site visits
- Days 31 to 60: the decisions only you can make
- Days 61 to 90: one cadence across every site
- What should you not do in the first 90 days?
- What if a site is not moving?
- What does month twelve look like?
You have three to seven sites, one of which is clearly the problem and two of which are quietly worse. You have a functional team that reports to you on paper and a set of site leaders who have their own reporting line to a general manager or a business unit president. You were promoted into this because you ran a plant well, and in your first week you have discovered that almost none of what made you good at that transfers directly.
That is the normal condition, and the discomfort is diagnostic rather than personal. The VP of Operations job is not a bigger version of the plant manager job. It is a different job that happens to be adjacent to it, and the first ninety days are mostly about establishing which of your instincts to keep. The plant-level version of this transition, for someone stepping into a single site, is in the four kinds of first 90 days.
The job you were actually given
Your job is to allocate scarce operating capability across sites and to make the decisions that no individual site leader can make from inside their own site. That is the whole of it, and everything else you do is either in service of that or is a plant manager’s job you have not yet let go of.
Two categories of decision belong only to you. The first is allocation: where capital goes, where your best people go, which site gets the attention this quarter and which one is told to hold. No site leader can make that call because no site leader can see the others clearly, and the ones who claim they can are usually advocating.
The second is arbitration between sites and functions. Sales wants a lead time your worst site cannot hold. Finance wants an inventory number that requires a reliability your network does not have. Two sites are quoting the same customer differently. Every one of those conflicts is unresolvable at site level by design, and if they are not resolved at your level they get resolved by whoever is loudest, which is a form of allocation nobody chose.
What is not your job, and this is the hard part, is fixing plants. You will be very good at fixing plants. Doing it personally in your first quarter feels productive, is genuinely productive at one site, and costs you the six weeks in which you were supposed to be establishing how the other five get run.
There is a fourth category worth naming, which is standard setting. You do not decide how a site schedules, but you do decide what every site is expected to be able to answer and how quickly. That is a genuinely different act from making the decision yourself, and confusing the two is the most common way a capable operations executive turns five site leaders into five deputies. Set the standard, publish it, hold people to it, and leave the method alone. The sites that meet the standard by a route you would not have chosen are the ones teaching you something.
Why the plant manager’s playbook fails at this altitude
The two-week plant diagnostic that works so well inside one site fails when applied to a network, and it fails in a specific way that is worth understanding before you try it. Run it at five sites and you will produce five accurate diagnoses that do not combine into anything.
The mechanism is that site-level diagnostics are designed to find the local constraint, and the local constraint at each site will be different. Site A has a maintenance problem. Site B has a scheduling problem. Site C has a decision latency problem. All three findings are correct. None of them tells you whether the network’s problem is that you have three sites making overlapping products, or that your capital has gone to the site with the most persuasive leader for four years running, or that your best three engineers have all been assigned to the same fire.
The second failure is temporal. The diagnostic assumes you can count things yourself, and you cannot count five sites yourself. Any attempt produces either a superficial pass at all five or a deep pass at one, and the deep pass at one is the trap, because it is genuinely useful work that quietly makes you that site’s plant manager.
What transfers is the instinct rather than the procedure. Count something yourself, distrust reporting that has not been checked, measure at the frequency of the decision. What does not transfer is the idea that your job is to find the constraint. At this altitude your job is to find the pattern across constraints, which is a different search.
The third failure is political and it is the one nobody warns you about. Running a full diagnostic at a site implicitly evaluates its leader, and doing it at all five in your first month tells every one of them that they are being assessed before they have been given anything. You need those people, several of them are probably good, and the ones who are good will read a month of unannounced auditing exactly as it appears. If you are going to look hard at a site, say why beforehand and say what you will do with what you find.
What should you measure first?
Measure the same three things at every site, in the same way, before you measure anything deeply. Output per labor hour on each site’s constraint, median time for an escalated decision to return, and gross margin by product family. Comparability across sites matters more than depth at any one of them.
The reason comparability beats depth is that a network problem is only visible in the variance. One site with a decision return time of two days and another at eleven tells you something no amount of study at either site alone would have revealed, and it tells you within a fortnight.
Expect the numbers not to be comparable when you first ask. Sites define their constraint differently, calculate labor hours differently, and allocate overhead differently, and each one has a reasonable explanation for its own method. Do not accept the explanations and do not launch a standardization project. Define the three measures yourself, in one paragraph each, and ask every site to produce them your way for one month alongside whatever they normally produce.
The objection you will get is that this is duplicative work during a busy quarter. It is. It is also the only way to get a network view inside ninety days, and the alternative is a standardization initiative that delivers in month nine.
One more measurement worth taking early, and it costs nothing. Ask each site how long it takes them to close a month and how much of that time is spent reconciling rather than reporting. The variance in that answer across a network is usually large, and a site spending a week reconciling is a site whose numbers you should be slower to trust and whose finance support you should be quicker to fix. It is also the least threatening question in the set, which makes it a good one to open with.
Which sites are actually yours to fix?
Fewer than you think, and probably one. A VP of Operations with five sites has the capacity to drive a genuine turnaround at one of them in a year while holding the others steady. Choosing which one, deliberately and early, is the highest-value decision of your first quarter.
The instinct is to pick the worst performer. That is frequently wrong. The right choice is the site where your attention produces the most change, which is a function of three things: how much of the network’s total margin sits there, whether the site leader can execute if unblocked, and whether the problem is one you can actually solve from your seat.
A badly performing site with a capable leader and a resource constraint is the best candidate you will ever get, because unblocking it costs you a decision rather than a year. A badly performing site with a leader who cannot execute is a much longer piece of work and it should be sequenced second, not first, regardless of how bad the numbers look.
Write the choice down and tell your boss which site you picked and why. The value of writing it is that in month five, when a different site has a bad quarter and everyone wants your attention there, you will have a documented allocation decision rather than a preference to defend.
The related trap is the site that asks loudest. In every network there is a site leader who is unusually effective at bringing problems to your attention, and the natural consequence is that your attention flows there regardless of where it would do the most good. That is not a character flaw in them, it is a structural feature of having limited attention and unlimited requests. The defense is to make the allocation decision in writing before the requests start arriving, because a documented prior choice is much easier to hold than a preference formed under pressure.
How do you tell a site problem from a portfolio problem?
Look at whether the same symptom appears at more than one site. A problem that shows up once is a site problem and belongs to the site leader. A problem that shows up at three sites is a portfolio, system, or policy problem, and no amount of site-level effort will fix it because the cause is not there.
This distinction is the single most useful mental habit at this altitude and it is easy to apply. Margin compression at one site is that site’s mix or pricing. Margin compression at four sites is a corporate discount policy, a pass-through failure in the commercial process, or a product portfolio that has drifted, and the site leaders being held accountable for it are being held accountable for something they do not control.
The same logic works on capability. One site struggling to schedule is a scheduling problem. Four sites struggling to schedule is a planning system, a data quality issue, or a demand signal problem that originates upstream in the commercial organization.
The place this matters most is in what it protects your people from. Site leaders in networks routinely carry accountability for cross-functional failures, and the resulting pattern is that capable operators burn out defending numbers they cannot move. The Harvard Business Review work on cross-silo leadership frames this as a collaboration challenge. From an operations seat it is more concrete than that: it is an accountability misassignment, and you are the first person in the structure with the standing to correct it.
One caution on applying the rule. Two sites showing the same symptom is genuinely ambiguous, and treating two as a pattern will send you looking for a system cause that is not there. Three is the threshold I use. Below it, ask both site leaders whether they think the cause is local, and take the answer seriously: operators are usually correct about whether a problem originates inside their own perimeter, even when they are wrong about what it is.
The last thing to establish before you act is what your predecessor was measured on, because the network has been optimizing toward it for years and the residue is still running. Ask your boss directly and ask two site leaders separately, and expect the answers to differ. Where they differ, you have found something more useful than either answer: the gap between what the organization was formally measured on and what it actually believed it was being judged on, and that gap explains a surprising amount of behavior that otherwise looks irrational.
From diagnosis to action
Everything above is establishing. What follows is the ninety-day sequence itself, and it is built around one constraint: you have roughly twelve site-weeks of your own attention in a quarter and you will be asked for forty. The plan below spends them deliberately rather than reactively.
A note on pace. Ninety days is one visit cycle, one decision cycle, and one cadence installation, and that is genuinely all it holds. Executives who plan four workstreams into a first quarter at this altitude are planning against a calendar that does not exist, because roughly a third of your time will be consumed by things that arrive on their own schedule: a customer escalation, a safety event, a budget cycle, a departure. Build the plan around twelve usable site-weeks and treat anything beyond that as upside.
Days 1 to 30: the site visits
Visit every site in the first thirty days, two days each, and run the same agenda at all of them. The consistency is the point. You are not diagnosing sites individually, you are collecting a comparable set of readings so that the variance becomes visible.
Day one at each site: walk the floor alone before any meeting, then count the site’s constraint for one hour yourself. Day two: sit with the site leader and go through the three measures, then ask three questions. What decision are you waiting on. What would you do with one more good engineer. And what are you being held accountable for that you cannot control.
That third question is the one that pays. Ask it at every site and the answers will overlap, and the overlap is your first portfolio finding. When four site leaders name the same uncontrollable, you have found something that no site visit could have surfaced on its own.
Resist fixing anything during the visits. You will see things you could resolve in an afternoon, and resolving them establishes that the way to get a decision at this company is to wait for the VP to be in the building. Write them down, take them into day 31.
A note on the visit agenda itself. Keep it to two days and hold the line on that, including at the site that is obviously worst. The temptation to stay a third day where the problems are visible is strong and it breaks the comparability you came for, because a three-day reading and a two-day reading are not the same instrument. Come back to that site in month two with a purpose rather than extending the diagnostic visit, and you will get better information both times.
Also decide in advance what you will do with what people tell you in confidence during these visits, and say it out loud at the start of each conversation. Site visits generate candid disclosures about other functions, about the previous VP, and occasionally about peers. If you have not decided your rule before the first visit, you will improvise it during the third one, and the sites talk to each other.
Days 31 to 60: the decisions only you can make
The second month is where you make the allocation and arbitration calls that were waiting for someone at your altitude, and you make them fast rather than well, because a backlog of unmade cross-site decisions is the most expensive thing in your inheritance.
Start with the list of blocked decisions collected during the visits. Sort by how long each has been waiting rather than by size. Then work the list and answer every one inside the month, including the answers that are no. A fast no lets a site leader stop waiting and start on the alternative, and the accumulated cost of people waiting is almost always larger than the cost of any individual wrong call.
Then make the allocation choice public. Name the site getting your attention this year, name what the others are expected to hold, and say it in a forum where all the site leaders hear it at once. The instinct is to avoid this because it looks like favoritism. The alternative is that everyone assumes they are competing for your attention, which produces exactly the advocacy behavior you do not want.
Last, take the overlap from question three and fix one of them. One cross-functional accountability misassignment, corrected, in your second month. It will be the thing your site leaders remember about your first year.
Expect resistance to the public allocation, mostly from the sites not chosen, and expect it to be framed as a concern about fairness. Answer it with the reasoning rather than with reassurance. A site leader told that their operation is expected to hold steady this year, and told exactly why, and given a date when the allocation will be revisited, is in a considerably better position than one left to guess. The version that damages you is the unstated allocation, where everyone can see where your calendar goes and nobody has been told the rule.
Days 61 to 90: one cadence across every site
The third month installs a single operating rhythm across the network, and its purpose is to make the second month unnecessary next quarter. A cadence is a set of standing questions asked at a fixed frequency by a named person with the authority to act.
Weekly, each site reports the same three measures in the same format. Not a deck. Three numbers and a note on anything blocked. Monthly, you run one review across all sites together rather than site by site, because the comparative view is the entire reason your role exists and site-by-site reviews reproduce the isolation you are trying to remove.
Inside the monthly, the standing question is the one that keeps the decision layer honest: what did not get an answer within a week, and from whom. Drop it and the queue rebuilds inside two quarters.
Quarterly, ask whether the allocation is still right. Not whether the chosen site is improving, which is a different question, but whether it is still the site where your attention buys the most change. Sometimes it is not, and changing the answer deliberately is very different from drifting toward whoever had the worst month.
One practical detail about the weekly report. Specify the format tightly and then never accept a variant, because the entire value is comparability and the first exception destroys it. Sites will propose improvements to the format, several of which will be genuinely better for that site, and accepting even one of them means you are back to five reports that cannot be laid side by side. Take the improvements into a scheduled annual revision, applied to all sites at once, and hold the format flat in between.
One more habit to install while the cadence is new. Require every site to bring one number that got worse alongside whatever improved. Reviews at this altitude drift toward good news faster than site-level reviews do, because each leader is presenting to a peer group rather than to their own team, and the social cost of a bad number in that room is higher. Making the bad number mandatory removes the choice and therefore removes the cost, and it is the difference between a review that surfaces problems and one that ratifies them.
What should you not do in the first 90 days?
Do not reorganize, do not standardize, and do not personally run a site turnaround. All three feel like leadership and all three consume the quarter you need for establishing. Reorganization before you understand the network rearranges a problem you have not yet located.
The standardization one deserves particular caution, because it is the most common first-quarter initiative at this altitude and it is almost always premature. Standardizing processes across sites is genuinely valuable and it should follow a diagnosis rather than substitute for one. Launched in month two, it produces a year of work aimed at differences that may not be causing anything.
And do not fire anyone in the first ninety days on the strength of a bad number. Site performance in a network is a joint product of the site leader and the allocation decisions they have been living under, and you have not yet been the one making those decisions. Judge people on a period in which they were operating under your allocation, not your predecessor’s.
And do not take over a function that reports to somebody else, however badly it is being run. New operations executives frequently absorb planning, or quality, or a piece of engineering, on the reasonable grounds that it is failing and they can fix it. The absorption is usually correct on the merits and almost always wrong on the timing, because you spend political capital in your first quarter on a structural change that will be read as empire building by exactly the peers whose cooperation you need for the cross-functional problems you identified in month one.
What if a site is not moving?
Establish first whether the site leader is blocked or unable, because the two look identical from your seat and require opposite responses. A blocked leader has named a decision, a resource, or an accountability problem and is waiting. An unable leader has not named anything.
The test is direct. Ask what they would do with a specific unblocking, and listen to how quickly and how concretely the answer arrives. A capable leader who has been blocked for eighteen months will answer immediately and in detail, because they have been rehearsing it. Someone who needs a week to come back with a plan is telling you something about capability rather than about blockage.
If the answer is blockage, unblock it and hold them to the result they described. If the answer is capability, the response is coaching and a defined check-in horizon, not a reassignment made in your first quarter on evidence collected under someone else’s allocation regime.
There is a third case that people miss, which is the site leader who is neither blocked nor unable but is optimizing for something other than what you are measuring. This usually traces to an incentive that was set years ago and never revisited, or to a promise made by a predecessor. The signal is that their numbers are stable and defensible while the behavior underneath them makes no sense against your priorities. That is the cheapest of the three to fix, because it is a conversation and a metric change rather than a capability problem.
What does month twelve look like?
The test is whether the network runs comparably without you assembling it. Every site reports the same three measures the same way, the monthly cross-site review happens whether or not you chair it, and blocked decisions surface within a week rather than a quarter.
Add one more test that is specific to this altitude. Have your site leaders started raising problems that belong to other sites? A network where each leader defends only their own perimeter is still a set of plants under common ownership. A network where a site leader says the constraint is at a different site is an operation, and the difference is the entire value your role adds.
Watch one leading indicator on the way there. Track how many of the decisions reaching you could have been made at site level under the thresholds you set, because that number is the cleanest read on whether the authority you granted is actually being used. A network where the thresholds were raised in month two and the escalation volume is unchanged in month eight has not devolved anything. It has published a policy, which is a different thing, and the gap between the two is where most operating model changes quietly die.
A limitation worth stating plainly. This ninety days assumes a network of three to seven sites with a shared product logic. Below three sites you are effectively a plant manager with a second location and the allocation problem is not yet real. Above ten, or across genuinely unrelated business lines, the comparability approach breaks down because your three measures will not mean the same thing at every site, and you need a segmented version of this that is beyond what one article can carry.
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.
The backlog of unmade cross-site decisions is usually larger than anyone has measured. Book a free Decision Velocity Audit and I will measure how long a decision takes to clear each of your sites, and where the variance is.
See also: What Kind of Employee Are You? 3-Word Test, Why Doing More Work Doesn’t Get You a Raise