Count how many times one number gets typed
Follow a single quote through a fabrication shop for a minute. Somebody builds it in the estimating workbook. It gets accepted, so somebody types it into the sales order template. Production needs it, so somebody types it into the scheduling board, or onto the whiteboard, or into a second spreadsheet that mostly matches the first one. The floor needs labels, so somebody types the piece marks into the label program. The job ships, so somebody types the whole thing one more time into the invoice.
That is five entries of the same information, by hand, by busy people, into systems that do not talk to each other. Every one of those entries is a chance to fat-finger a number, grab the wrong revision, or skip a line. And most shops I have been in do some version of this on every single job.
What does that actually cost? Not in frustration. In dollars.
The government measured it
In 2004, the National Institute of Standards and Technology published a study called Cost Analysis of Inadequate Interoperability in the U.S. Capital Facilities Industry (NIST GCR 04-867). "Inadequate interoperability" is a government phrase for a plain problem: your software systems, spreadsheets, and paper cannot pass information to each other, so your people re-enter it by hand and fix the mess when it goes wrong.
NIST put the total cost at $15.8 billion a year. That number covers the whole capital facilities industry, but the part that matters for this article is the slice they measured for specialty fabricators and suppliers: $2.2 billion a year, against $177.2 billion in annual receipts. A follow-up analysis by NIST economist Robert Chapman worked that out as a share of revenue for each group, and for fabricators it came to 1.2% of annual receipts, the highest rate of any group actually building the work.
Sit with that. For a fabrication shop doing $14 million a year, 1.2% is about $167,000. Every year. That is our arithmetic, not NIST's, but it is simple multiplication on their numbers. Pick your own revenue and run it: 1.2% of whatever your shop bills.
And mind you, those are 2002 dollars. Twenty-some years of inflation later, the real number is well north of that. The problem has not gotten smaller either. A 2021 study by Autodesk and FMI estimated that bad data, meaning data that is inaccurate, incomplete, or stuck where nobody can use it, may have cost the global construction industry $1.85 trillion in 2020, including $88.7 billion in avoidable rework.
Two more findings from the NIST work are worth pulling out, because they point straight at the fix. First, manual re-entry alone cost fabricators roughly $244 million a year: typing the same data into system after system. Second, and this is the one that got me, nearly 85% of the fabricators' total burden came from what the study calls inefficient business process management. Not exotic technology failures. The everyday overhead of managing jobs, documents, and information requests across systems that do not connect. That was the highest share of any stakeholder group in the study, by a wide margin.
In other words: the researchers found that fabricators lose more of their revenue to disconnected information than the architects upstream or the owners downstream. The people actually making the things pay the steepest data tax.
The spreadsheet grandmaster
Here is where I tell you something you already know.
Somewhere in your shop there is a workbook. Fifteen tabs, maybe twenty. Lookup tables feeding formulas feeding a summary sheet, a macro somebody wrote years ago, cells you are not allowed to touch highlighted in a color whose meaning has been lost to time. It prices your work, or schedules it, or both. And exactly one person knows how it actually works.
That person is the spreadsheet grandmaster. Every shop has one. When they are in the building, quotes go out and the board stays current. When they take a week at the beach, the whole shop holds its breath. Somebody copies last month's file, changes what they think needs changing, and hopes. When a formula quietly breaks three columns deep, nobody catches it until a job prices wrong, and you find out about it from the customer or from your margin report, whichever hurts more.
I am not making fun of the grandmaster. I have been the grandmaster. When I was running precast operations, I built those workbooks myself, and I was proud of them, because they were genuinely the best tool we had. But a business that cannot let its sharpest person retire, or even take a vacation, without risk is not running on software. It is running on a person, and people wear out.
The NIST study never uses the word "grandmaster," but it is measuring that shop. Copies of copies. One number living in five places. Knowledge in one head instead of in a system. The 1.2% is what that costs when you add it up across a year.
Where the money actually goes
Nobody writes a check for $167,000 labeled "re-keying." The leak hides in small places, which is why it survives.
It goes out through a mistyped elevation that gets fabricated before anyone catches it, and the remake ships on your dime. Through the crew building from revision B while the customer approved revision C, which is the expensive way to find out your document control lives in an email inbox. Through hours of skilled office staff typing the same job into a fourth system, hours you paid for once and got nothing new from. Through the office being a day behind the floor, so the answer to "when will it ship" starts with somebody walking out to go look. Through a broken formula in the quoting workbook that prices work below cost for a month before the margin report tattles.
Any one of those sounds small. NIST's contribution was to add them up and show that they are not small. They are 1.2% of everything you bill.
The fix is structural, not heroic
You cannot fix this by telling people to type more carefully. The error rate is not a character flaw. It is a property of the system, and the system is the problem: the same fact stored in five places has to be maintained in five places, and it will drift.
The fix is a relational database. That phrase sounds technical, but the idea is plain: every fact gets stored exactly once, and everything that needs the fact refers to it instead of copying it. In a fabrication shop, the natural center is the approved shop drawing, because that is the document your customer signed off on and your crew builds from. Store the job there, once. Then the quote, the sales order, the production schedule, the piece labels, the bill of materials, and the invoice are not five separate documents somebody maintains. They are five views of the same record. Change the record and every view is already right. There is nothing to re-key, because there is no second copy to key into.
The other half is a mobile app, and it matters just as much. The office being a day behind the floor is a data-entry problem too: status lives in the lead man's head until somebody walks it back to a desk. Put the system on a tablet at the saw, the weld bay, the finish room, or the yard, and the update happens where the work happens, at the moment it happens. One tap when the piece is done. The office sees it now, not tomorrow. Ask your scheduler what they would give for that.
This is exactly what an ERP built for fabrication is: one relational database wrapped around your actual workflow, with the shop drawing as the record and the floor holding up its end from a tablet. It is the difference between software that adds another place to type and software that removes the typing.
Why this is bigger than the money
Say a shop plugs the leak. What does it actually buy?
Recovering 1.2% of revenue matters on its own; in a business running single-digit margins, that can be a fifth of the profit. But the second-order effects are the real story. A shop that is not bleeding clerical hours can quote more work without hiring more office staff. A shop whose knowledge lives in a system instead of in one head can open a second location without cloning the grandmaster. It can promote the grandmaster, who is usually the smartest person in the building, to work worth their brain. It can train a new hire in days instead of years, because the system carries the process. It can give raises out of margin it used to donate to typos.
And here is the part I care about most, being from East Tennessee and watching what happens when shops grow or shrink. Fabrication shops are anchor employers. When a precast plant or a metal fab shop or a millwork house scales up, it hires welders, finishers, drivers, estimators. Those paychecks stay in the county. The 1.2% is not just leaking out of one P&L. Multiplied across every shop running on disconnected spreadsheets, it is payroll, expansions, and second locations that never happen. Plugging it is economic development you can do without a ribbon cutting.
The shops that fix their data problem do not just keep more of what they earn. They get room to grow, and the towns around them grow too.
Where to start
Start by counting. Take one recent job and tally how many times the same information got entered by hand between the first phone call and the final invoice. If the answer is one, close this tab and go on about your day; you have my respect. If the answer is five, you now know where your 1.2% lives.
We build DhyanaERP for exactly this problem: one system for fabrication shops where quoting, shop drawing approval, production tracking, delivery, and invoicing all run off one record, with a mobile app the floor actually uses. The Precast Edition is live and running real production today, and we are building Metal, Millwork, and Glass editions on the same core with early access shops right now. If your shop runs on a grandmaster and a stack of workbooks, come talk to us. Bring the workbook. We have seen worse.
References
- Gallaher, M. P., O'Connor, A. C., Dettbarn, J. L., Jr., & Gilday, L. T. (2004). Cost Analysis of Inadequate Interoperability in the U.S. Capital Facilities Industry (NIST GCR 04-867). National Institute of Standards and Technology. https://nvlpubs.nist.gov/nistpubs/gcr/2004/NIST.GCR.04-867.pdf
- Chapman, R. E. (2005). Inadequate Interoperability: A Closer Look at the Costs. 22nd International Symposium on Automation and Robotics in Construction (ISARC 2005). Source of the per-stakeholder figures: $2.2 billion for specialty fabricators and suppliers against $177.2 billion in receipts (1.2%), manual re-entry costs of $128M and $116M in the design and construction phases, and the 84.6% business-process share. https://www.iaarc.org/publications/fulltext/isarc2005-85chapman.pdf
- Autodesk & FMI. (2021). Harnessing the Data Advantage in Construction. Estimated $1.85 trillion global cost of bad data in 2020, including $88.69 billion in rework. https://www.autodesk.com/blogs/construction/autodesk-fmi-study-global-construction-industry-data-strategies/
The $167,000 figure is our arithmetic: 1.2% (the NIST-derived inefficiency rate for specialty fabricators) applied to a hypothetical shop with $14 million in annual revenue, in the study's 2002 dollars. Your number scales with your revenue.