One of the biggest advantages of modular and offsite construction is that the factory is a measurable environment. Same place, same equipment, same people, every day. You can see where time goes and where material goes in a way that is almost impossible on a job site.
So why do so many offsite operations still run on gut feel?
Most plants measure output. Modules or panels per week is on the whiteboard. Far fewer measure what drives it: labor hours by station against a standard, first pass yield, rework hours and where they came from, material substitutions against the bill of materials. In the field it is thinner still. Set duration, punch items by cause and damage in transit often live in emails and photos, if they are captured at all.
The gap between factory data and field data
The most expensive problems live in the gap between those two data sets. A detail that is wrong in design, often because pre-production was rushed, gets built in the plant, ships, and shows up at the set as a field fix. The field fix gets charged to a site budget. The crew solves it, because crews always solve it. The lesson never makes it back to the designer or the station that built it, so the next unit carries the same problem.
Every good set crew has a little MacGyver in it. Hand them a bad detail, a roll of tape, some caulking and maybe a sawzall, and they will make it work. That is a strength and a problem. Because the fix works, nobody reports it, and the designer and the plant never learn it was needed. I have asked for detailed tickets on significant job sites where the manufacturer owned the set crew, and I never saw data clear enough to trace an issue back to its root cause and fix it for good.
It gets worse when rework is filed under another name. Call it R&D, call it a change order, call it warranty, and it disappears from view. You cannot fix what you file under the wrong code or worse a code used to make performance look better.
Rework hits every project, from single family to mid-rise
Rework is not a modular problem or a site-built problem. It shows up at every size and type of project, and the numbers are not small.
- Single family homes: the public homebuilders tracked by Warranty Week set aside an average of $2,980 for warranty work on every home sold in 2024, and paid out $1.07 billion in warranty claims that year. That is money spent fixing homes after the buyer has moved in.
- Across project types: research published in the ASCE Journal of Construction Engineering and Management found that direct rework costs often reach about 5% of total construction cost. That is somewhere between 25-50% of what many contractors hope to make on a project.
- Industry wide: FMI and PlanGrid estimated that poor project data and miscommunication caused $31.3 billion of rework in the U.S. in 2018, 48% of all rework. The same survey found construction professionals spent about 14 hours a week, 35% of their time, looking for project information, resolving conflicts and fixing mistakes.
- Multifamily, mid-rise and larger: when rework turns into a dispute, the cost climbs fast. Now the fix carries lawyers, experts and months of delay on top of it, and the relationship takes a hit too.
The common thread at every scale is information. Most rework traces back to something someone knew, or should have known, before the work was done. Capturing that information and putting it to use is exactly where a factory has the advantage, if it chooses to use it.
What to measure in a modular factory and in the field
This does not take a big software rollout. It takes consistency. Track labor hours by station against a standard. Log every rework event with its root cause and where it originated: design, factory, transport or site. Track bill of materials variance and every substitution. Record shipping and handling damage. Code field punch items by origin. Then close the loop every week, with design, production and field in the same room looking at the same numbers. What are we tracking? Cost of labor, materials, equipment, and delay.
Underneath all of it, keep one source of truth for project data, so the drawings, the BOM, the cut lists and the machines are all working from the same model. Most errors I have seen trace back to two versions of the truth.
Will good data uncover uncomfortable truths? Absolutely. That is the point. The cost of finding a problem in the data is a conversation. The cost of finding it repeated across dozens of modules at the set is a crane on standby, a crew on overtime and a customer losing confidence.
Mike Tyson said it best: "Everybody has a plan until they get punched in the mouth." Good data is how you see the punch coming before the crane shows up.
Risk does not hide in what you measure. It hides in what you don't.