Bad data didn't cost construction $1.85 trillion. Bad plans did.

Bad data cost $1.85 trillion in a single year

1.85 trillion dollars in 2020, by the broadest estimate available. Autodesk and FMI surveyed more than 3,900 construction professionals worldwide for the 2021 report Harnessing the Data Advantage in Construction and priced what the industry loses to data that is inaccurate, incomplete, inaccessible, inconsistent or late. Almost a third of respondents said more than half of their project data is bad, and that it leads to poor decisions more than half the time.

The industry's reflex is to treat this as an IT problem: buy a new system, standardise the formats, get everything in one place. But data does not spoil on its own. Ask where an inaccurate quantity or a contradictory requirement first entered the record, and the answer is rarely a database. It is a document that somebody wrote.

Bad data comes from the plans

From the plans, mostly. When PlanGrid and FMI surveyed nearly 600 US construction leaders for Construction Disconnected in 2018, they found that poor project data and miscommunication cause 48 percent of all rework on American jobsites: 31.3 billion dollars a year. The same study clocked project staff at more than 14 hours per week of non-productive work, about 5.5 of them just looking for project data and information.

Rework is where bad data becomes visible. It is rarely where the data went bad. Ruben Vrijhoef and Lauri Koskela put the pattern in one sentence in 2000, and supply chain research has kept confirming it since:

Most [waste and problems] are caused in another stage of the construction supply chain than when detected.

In construction, the stage where the defects are born is usually the earliest one: design and pre-construction. A tolerance that contradicts itself between drawing and specification. A quantity carried over from a template project. A code reference that changed after the section was written. Each is one flawed line in a tender package. Each becomes bad data in every estimate, schedule and system that inherits it.

One ambiguous clause multiplies through every handoff

Because nobody downstream can refuse it. A project is a chain of handoffs: the client's requirements go to the designers, the design goes into a tender package, and the winning contractor passes slices of it to dozens of subcontractors. Every handoff is a re-interpretation. The estimator reads the ambiguous clause as a risk and prices it. The site team reads it as an instruction and guesses. The guess becomes a query, then a change order, sometimes a claim. One hour of drafting ambiguity turns into hundreds of hours of downstream repair.

The UK's Get It Right Initiative has measured the shape of that bill: error costs around 5 percent of project value directly, and 10 to 25 percent once indirect and latent costs are counted. Its field research also explains why the problem persists: "the majority of the initial cost of error is incurred by Tier 2 Contractors and this cost is often invisible both to Tier 1 Contractors and to end Clients." The designer who wrote the clause never sees the invoice. So the same defect ships again on the next project.

Fix the documents, and the data follows

The disputes data closes the loop. Arcadis's fifteenth annual Global Construction Disputes Report (2025) names errors and omissions in the contract document as the leading cause of construction disputes for the third year in a row. A dispute is the last stop on the line. It started years earlier, as a sentence that could be read two ways.

The diagnosis decides the remedy. If the problem is bad data, the remedy is better plumbing: systems, formats, one place for everything. Useful, but downstream. It cleans the lake while the river keeps carrying silt. If the problem is document quality at the source, the remedy is to check the plans before they are released: every drawing against every specification, every reference against the current rulebook, every quantity against the design it claims to describe. At the scale of a modern tender package, that reading work has been impossible to do by hand. It has stopped being impossible.