Most DFW manufacturers run this decision backwards. They price a provider against a salary, hire, and then find out one person cannot cover a two-shift plant. Start with your production calendar and your control network instead. Your own operating profile picks the model, and for most metroplex shops running under 200 people with a second shift already on the calendar, that answer lands somewhere between the two extremes rather than at either one of them.
A DFW manufacturer should keep IT in-house when one site runs one shift with no controlled data, and should move to an MSP or a co-managed model once production hours exceed what one person can physically cover.
That sounds obvious written down. It is not how the decision gets made.
Here is what normally happens. A plant manager pulls a quote for manufacturing IT services in Dallas, divides it by twelve, and compares it to a salary on Indeed. Hiring looks cheaper. So they hire. Nine months later that same plant is running a job posting it cannot fill, and the one person they did hire is answering a call at 6 in the morning because a press cell dropped off the network.
The math was never wrong. It was measuring the wrong thing.
I have sat in that conference room. More than once. The number that decides this is not the salary. It is the gap between the hours your plant runs and the hours a human being is actually available.
What this decision actually turns on inside a plant
In-house IT means you employ the people who support your systems. An MSP means you contract that support to an outside firm on a fixed monthly agreement. Co-managed means you do both. Your employee owns what happens on the floor, the provider carries the after-hours load, the security stack and the patching discipline, and between them you get the depth a single headcount could never justify on its own.
Office businesses can pick any of the three and survive a bad choice. A plant cannot. That is the whole difference.
When a domain controller dies at an accounting firm, invoices go out late. Annoying, not fatal. When the same thing happens at a fabricator running a second shift, a scheduled job sits in a queue nobody can release. The machines idle. The operators clock in anyway. You pay for both.
Siemens and Senseye put a number on that difference. In their 2024 study of unplanned downtime, the world’s 500 largest companies lose roughly $1.4 trillion a year to downtime, about 11% of total revenue, and an idle automotive line runs to $2.3 million an hour at the top end. Their surveyed plants averaged 27 hours of unplanned downtime a month. That is down from 39 hours in 2019, so the trend is good. Your shop is not a Global 500 automaker. The ratio still travels.
So the real question changes. It is not which model is cheaper. It is which model keeps hours out of that column.
The DFW hiring market decides this before your budget does
Here is the part almost nobody checks first. Before you compare a salary to a monthly fee, you have to be able to hire the person. In Dallas-Fort Worth that depends heavily on which role you actually need. The two are not close.
These are the current federal wage estimates for the metroplex, from the Bureau of Labor Statistics Occupational Employment and Wage Statistics program for the Dallas-Fort Worth-Arlington MSA, May 2025 reference period. The loaded column applies the 1.43x factor from the BLS Employer Costs for Employee Compensation release for the first quarter of 2026, where total compensation averaged $46.60 an hour against $32.60 in wages.
| Role | Employed in DFW | Annual mean | Loaded at 1.43x |
|---|---|---|---|
| Computer user support specialist | 22,600 | $63,030 | $90,133 |
| Computer network support specialist | 3,340 | $80,300 | $114,829 |
| Network architect | 11,740 | $108,020 | $154,469 |
| Database administrator | 2,430 | $118,560 | $169,541 |
| Information security analyst | 7,080 | $133,790 | $191,320 |
| Computer and information systems manager | 30,690 | $185,720 | $265,580 |
The role you actually need barely exists here
Look at the second column, not the third. DFW employs 22,600 user support specialists. It employs 3,340 network support specialists. One seventh the size. That gap matters.
Network support is closer to what a plant needs than help desk is. Password resets are not the problem in a machine shop. The switch stack feeding the cell controllers is. You are fishing in the smaller pond, and every other manufacturer in the metroplex is fishing there too.
Now stack a security requirement on top. An information security analyst carries a $133,790 mean in this market. Loaded, that is $191,320. Nobody at a 120-person fabricator is putting that seat in the budget. Not one. So plant security work ends up either outsourced or quietly not done, and the second option is far more common than anybody says out loud.
That is the recruiting reality behind the decision. You are not choosing between a person and a provider. You are choosing between a person you may not be able to hire and a provider who already employs six of them.
Why the salary number misleads
The wage figure is not the cost. Not close. Employer payroll taxes, health coverage, retirement match, paid leave and insurance together push a $63,030 salary to about $90,133 in real annual expense, which is the figure your CFO already uses for every other role in the building. And that is before you buy the person a laptop, a phone, a remote access tool or a single certification.
Apply that adjustment to every quote comparison you run. Skip it and the comparison is not a comparison. If you want the full dollar-for-dollar version of the exercise, we walked through it in our breakdown of in-house IT versus an outsourced MSP.
Coverage arithmetic, the math that breaks in-house first
This is the calculation I wish more plant leaders ran before posting a job. It takes four minutes. It usually ends the debate.
One full-time employee is paid for 2,080 hours a year. Take out three weeks of paid time off and ten holidays. Now you are at roughly 1,880. Take out training, vendor calls, and the days that person is genuinely unavailable. About 1,840 hours of real coverage is left. Call it 1,840 to be generous.
Now count your production hours.
| Plant schedule | Production hours per year | FTEs to cover it | What one hire actually covers |
|---|---|---|---|
| One shift, 5 days | 2,080 | 1.1 | 88% |
| One shift plus Saturday | 2,500 | 1.4 | 74% |
| Two shifts, 5 days | 4,160 | 2.3 | 44% |
| Two shifts plus Saturday | 5,400 | 2.9 | 34% |
| Three shifts, 6 days | 7,488 | 4.1 | 25% |
Take a two-shift plant running Saturdays. That is about 5,400 production hours a year. One hire covers 1,840 of them. Call it 34%. And that assumes the person never sleeps through a call.
To cover that calendar properly you need close to three people. Three user support hires at the DFW loaded mean is roughly $270,000 a year. It still buys you no security analyst. No network architect. Nobody who has configured a firewall rule for an OEM tunnel before. That is the number to put next to a monthly agreement, and it is the number nobody puts there.
There is a quieter version of this problem. One person means one point of failure. Your single IT hire takes a week in Colorado. Coverage that week is zero. Not reduced. Zero. The operators find out at 5 in the morning on a Tuesday.
Six questions that pick the model for you
Answer these about your own operation. Not about vendors. Not about pricing. The answers point at a model before anyone quotes anything.
- How many shifts does the plant run, and does anything run on Saturday?
- How many devices sit on the control network that cannot take an unplanned reboot?
- When ERP or MES stops, does the line stop with it, or do you keep running on paper?
- Do ITAR-controlled or CUI drawings live anywhere a support technician can reach?
- How many physical sites are you supporting, and how far apart are they?
- What does one hour of unplanned line-down actually cost you, in labor plus lost throughput plus expedited freight?
Question 6 is the one most plants have never calculated. Do it once. Say an hour of downtime costs you $8,000, and say you take 27 hours of it a year. That is $216,000. Annually. On something a support model is supposed to be preventing. Every quote in front of you suddenly reads differently.
Question 4 changes the answer more than any of the others. If controlled technical data touches your network, the people who can support that data are limited by regulation, not preference. The obligations run deeper than most general providers expect. We covered that ground separately in our guide to ITAR, CMMC and downtime obligations for DFW manufacturers.
Reading your answers against the three models
Map what you wrote down against this. It is not a scoring rubric and it does not need to be. Most plants find their own row immediately.
| Plant profile | Best fit | Why |
|---|---|---|
| 1 site, 1 shift, under 60 users, hosted ERP, no controlled data | In-house generalist | Coverage gap is small and a single capable person genuinely fits the calendar |
| 1 site, 2 shifts, 60 to 150 users, on-prem ERP | Co-managed | One hire covers 44% of production hours, so nights and weekends need a bench |
| 2 or more sites, any shift pattern | MSP or co-managed | Travel time between plants eats the coverage a single employee could offer |
| ITAR or CUI drawings on the network | Co-managed with a vetted provider | Access control becomes a staffing question, not a technology question |
| Legacy HMIs on unsupported Windows, no OT inventory | MSP with OT experience | Segmentation and compensating controls are specialist labor you cannot hire part-time |
| Under 40 users, break-fix relationship already working, low downtime cost | Leave it alone | Nothing here justifies the change, and the switching cost is real |
Where in-house genuinely wins
Let me argue against my own industry for a minute. The honest version of this matters more than the sales version.
Picture a single-site plant. One shift, 45 users, vendor-hosted ERP, no controlled data. That company is well served by one good generalist. Full stop. That person learns the presses. They know which operator always unplugs the wrong cable and which cell controller needs a specific boot order. Institutional knowledge on a plant floor is worth more than most providers admit, and it does not transfer cleanly in a runbook.
We have told companies in exactly that shape not to hire us. Twice this year. Both of them were better off putting the money into a second person on their own payroll and buying a backup and monitoring stack their own team could run without us standing behind it.
The model breaks at a predictable place. Three events do it. A second shift starts. A second site opens. A customer sends a security questionnaire the generalist cannot answer. Watch for those three. They are your transition markers. They show up months before the pain does, which is the only reason watching for them is useful at all.
Why most DFW plants land in the middle
Co-managed is not a compromise. For a metroplex plant between 80 and 250 people, it is usually the only model that matches the actual work.
Your employee owns the floor. They know the equipment. They walk the line. They handle whatever needs hands on a device at 7 in the morning. The provider carries the hours your employee physically cannot be awake for, plus patching, monitoring, backup validation, and the security tooling that a single salary in this market would never stretch far enough to cover. Neither side does the other’s job. That is the point.
The pressure behind this is not theoretical. Manufacturing accounted for 27.7% of the incidents IBM X-Force observed in 2025, the fifth straight year the sector has topped that list, with vulnerability exploitation behind 40% of incidents overall. Manufacturing Dive’s coverage of the same index makes the reason plain. Operational urgency plus aging equipment is exactly the combination an extortion crew looks for.
Aging equipment is where a general IT approach stops working. NIST says so directly in SP 800-82 Revision 3, the federal guide to operational technology security, which notes that many standard security controls need modification before they apply to systems built for availability first. A PLC does not take a patch on Tuesday because Microsoft released one. Ever. If your provider does not already know that, they will learn it on your line. At your expense.
That is also why the vendor conversation is a separate exercise from the model conversation. Once you know which model fits, the evaluation criteria change. We laid those out in the 9 factors Texas manufacturers use to score providers. The OT and IT security boundary is worth reading alongside it.
Run these three numbers before you post the job
You do not need a consultant for this part. You need an afternoon and a spreadsheet.
- Your annual production hours, counted from the actual shift calendar rather than the posted one
- Your fully loaded cost per hire, meaning the DFW mean for the role multiplied by 1.43
- Your cost of one hour of unplanned line-down, labor plus throughput plus expedite
Put those three next to any monthly agreement you are considering. The comparison stops being about price. It starts being about coverage, which is the thing you were actually buying.
One more caution. Do not let a provider tell you the answer before you have run the numbers yourself. A firm that quotes you a model without ever asking about your shift calendar has not really looked at your plant, and the proposal on the table will show exactly that. Ask them the six questions above. See whether they already knew to ask you.
Where Uprite fits in this decision
We run manufacturing IT support across Texas. The DFW plants we work with are mostly machine shops, fabricators, and electronics and aerospace suppliers between 60 and 400 people. Most are on a co-managed arrangement in Dallas-Fort Worth where they keep an internal person and we carry the rest.
We are a poor fit for a 30-person shop with one shift and no compliance exposure. We are also the wrong call if you want someone to take full ownership and never speak to your team again. Plant IT does not work that way, and pretending otherwise sets both sides up to fail. If you want the broader metro picture first, our managed IT services in Dallas-Fort Worth page covers the general case.
If you would rather have somebody walk the floor and run the coverage math with you, that is what an assessment is for.
What DFW plant managers ask before they decide
Can one IT person really support a plant?
One person can support a single-site plant running one shift with roughly 60 users or fewer. Past that, a single hire covers less than half your production hours, and nights and weekends go uncovered. Coverage math decides this, not skill level.
Is an MSP more expensive than hiring in Dallas-Fort Worth?
Usually not, once loaded cost is applied. A $63,030 user support salary costs about $90,133 a year in DFW after taxes and benefits, and a two-shift plant needs close to three of those seats to cover its calendar properly.
What happens to our current IT person if we bring in a provider?
In a co-managed arrangement they stay, and the job usually gets better. They stop taking calls at 2 in the morning and start owning floor work, vendor coordination, and the equipment knowledge no outside firm can replicate. The provider absorbs the after-hours load, the patch cycle, the backup validation and the security tooling, which is almost always the part your internal person had been asking for help with anyway. Retention tends to improve, not drop.
Our controls vendor already handles the plant floor. Does that cover us?
Rarely. Controls integrators own the process and the programming. They do not typically own network segmentation, Windows patching, backup validation, remote access governance for your OEM vendors, or your response at 3 in the morning when somebody encrypts the file server holding every drawing you have.
How long does it take to switch from in-house to a co-managed model?
Plan on 60 to 90 days for a single DFW site running two shifts. Discovery and OT inventory eat the first three weeks.
We are not a defense supplier. Do we still need to worry about ITAR?
Possibly, because ITAR follows the technical data rather than the customer relationship. A DFW machine shop cutting a single part for a prime contractor can hold controlled drawings on its own file server without ever shipping anything overseas, and the registration obligation tends to catch those shops far later than it should.
What is the single strongest signal that in-house has stopped working?
A second shift starting. That one event moves your coverage from roughly 88% of production hours down to about 44% literally overnight, and no amount of goodwill or overtime from a single IT hire is going to close a gap that size.
