Rework is expensive, but published estimates vary widely. This article separates the numbers, explains what 'bad information' means on a construction project, and shows where earlier document coordination can reduce avoidable rework.
|
IN SHORT • Published rework percentages vary because researchers do not all define 'rework' the same way; that is why a single universal percentage is misleading. • Autodesk and FMI estimated that bad construction data may have contributed to $1.85 trillion in global industry costs in 2020 and about $88.69 billion in rework. • Academic case studies have reported direct rework costs in the low single digits of contract value, while newer contractor data suggests measured field rework can be lower when definitions are narrower. • Design errors, omissions, inconsistent information, communication failures and client changes repeatedly appear in rework research. • The most useful question is not 'What percentage does rework cost?' but 'Which rework causes can we move upstream, before work is installed?' |
$1.85TEstimated 2020 global cost impact of 'bad data' — Autodesk/FMI |
$88.69BEstimated rework associated with bad data in the same study |
0.38–0.76%Recent measured field-rework range in one contractor dataset reported by ASCE |
Construction has no shortage of dramatic rework statistics. Depending on the article, report or conference slide, rework may be described as 2%, 5%, 10% or even more of project cost.
The problem is that those numbers are often repeated without explaining what researchers counted.
Some studies include only direct labor and material needed to redo installed work. Others include change orders, post-completion corrections, supervision, disruption, schedule effects or indirect management cost. Some focus on building projects; others on infrastructure. Some measure actual cost records; others rely on surveys.
So the first rule when discussing rework is simple:
|
Rework is real and costly, but there is no single percentage that applies to every project. |
What the research actually says
One useful benchmark comes from research by Autodesk and FMI. Their 2021 study surveyed more than 3,900 construction professionals and defined “bad data” as project information that is inaccurate, incomplete, inaccessible, inconsistent or untimely. The study estimated that bad data may have contributed to $1.85 trillion in global construction industry costs in 2020. It also estimated $88.69 billion in rework associated with decisions made using bad data.
That figure is not a claim that drawing errors alone caused $88.69 billion of rework. The category is much broader. It includes poor information quality across construction workflows.
But the definition is useful because it describes what project teams see every day: the wrong information, missing information, conflicting information, late information or information that exists somewhere but cannot be found when needed.
Academic research has produced smaller and more project-specific rework numbers. A frequently cited study by Peter Love and Heng Li examined two Australian construction projects and reported rework costs of 3.15% and 2.40% of contract value. Errors and omissions in contract documentation, along with client- and end-user-driven changes, were among the primary causes.
More recently, ASCE summarized research based on actual contractor records in which precompletion field rework averaged 0.38% of contract value and rose to 0.76% when postcompletion corrections were included.
Those figures look very different, but they are not necessarily contradictory. They are measuring different populations and different definitions.
Why definitions matter more than headlines
Imagine a $50 million project where a trade installs a service using an outdated detail.
If the direct correction requires $40,000 of labor and materials, a narrow rework measure may record $40,000.
But the actual project impact can also include:
· Superintendent and project-manager time
· Trade resequencing
· Equipment standby
· Consultant review
· RFI processing
· Acceleration elsewhere in the schedule
· Procurement changes
· Delay exposure
· Commercial negotiation
Whether those secondary impacts are counted determines the rework percentage.
That is why companies trying to benchmark themselves should establish their own definition before comparing against industry numbers.
Where bad information enters the rework chain
The Autodesk/FMI definition of bad data is especially relevant to drawing review because the five failure modes map directly onto construction documentation.
Inaccurate information
A dimension, note, rating, equipment value or location is simply wrong.
Incomplete information
A detail, dimension, connection, clearance requirement or scope instruction is missing.
Inaccessible information
The correct information exists, but it is buried in a specification section, an old drawing issue, a schedule or a separate consultant package that the person making the decision cannot easily find.
Inconsistent information
Two project documents describe the same condition differently.
Untimely information
A revision arrives after pricing, procurement, fabrication or installation has already started.
These are not abstract data-governance problems. They are practical construction risks.
A drawing can be correct and still contribute to rework
One of the hardest coordination problems is that the error does not always live on a single sheet.
The architectural plan may correctly show a room.
The mechanical plan may correctly show a unit.
The structural plan may correctly show a column.
Yet when those three pieces of information are considered together, the column may occupy the service clearance required by the mechanical unit.
The problem exists at the interface.
This is why document quality cannot be measured only by whether individual sheets are internally correct. Coordination quality also depends on whether the package works as a connected set.
The cost curve changes when the issue moves downstream
A conflict discovered before construction may require a coordination meeting, consultant clarification and revised detail.
The same conflict discovered after material has been fabricated or installed can require demolition, replacement and schedule recovery.
The technical issue is identical. The project context is not.
That leads to a more useful way to think about rework prevention:
Stage 1: design review
The cost is mostly professional time.
Stage 2: preconstruction
The cost may include estimating changes, coordination and procurement adjustment.
Stage 3: fabrication
The cost can include material, shop drawing and supplier impacts.
Stage 4: installation
The cost can include demolition, labor, equipment, supervision and schedule disruption.
Stage 5: post-completion
The cost may expand into warranty, access, occupancy and reputation impacts.
The earlier a preventable information problem is resolved, the fewer downstream systems it can disturb.
What repeatedly causes rework?
The literature does not produce one universal ranking, but several themes recur.
Design and documentation problems
Research on contract documentation has repeatedly connected documentation quality with rework. Errors, omissions and inconsistencies can force site teams to interpret intent, request clarification or proceed based on assumptions.
Communication failures
A 260-project path-analysis study by Love and colleagues found project communication to be a statistically significant contributor to rework costs, alongside client-directed changes and site/subcontractor factors.
Client and design changes
Not all rework is preventable. Owners change their minds. Scope evolves. Site conditions differ from expectations. Design development continues.
A good review process will never eliminate all rework because some rework originates from legitimate change.
Information-management failures
Teams may possess the correct information but still act on the wrong version, fail to communicate the revision, or be unable to locate the requirement when it matters.
The practical target is avoidable rework
Construction leaders should be careful with claims such as “we can eliminate rework.” That is unrealistic.
A better objective is to reduce avoidable rework caused by information that could reasonably have been identified earlier.
That means asking:
· Did two issued drawings disagree?
· Was a required detail missing?
· Did a revision change one discipline without updating dependent information?
· Did the drawing and specification describe the same item differently?
· Was a trade conflict visible before fabrication or installation?
· Was the information available, but too difficult to find?
Those questions lead directly to process improvement.
A better KPI than raw rework percentage
If a contractor wants to improve, the total rework percentage alone is not enough.
Track the root cause.
A useful internal rework taxonomy might separate:
- Owner/client change 2. Design error 3. Design omission 4. Drawing/specification inconsistency 5. Coordination failure 6. Wrong document version 7. Fabrication error 8. Installation/workmanship error 9. Unforeseen site condition 10. Supplier/material issue
Then track when each issue was first discoverable and when it was actually discovered.
That gap is often where the biggest opportunity sits.
Where AI-assisted drawing review fits
AI drawing review should not be sold as a way to “eliminate rework.”
The credible opportunity is narrower and more useful: improve the probability that document-related problems are surfaced before they travel downstream.
A system can help compare drawing references, scan for inconsistent information, identify potential cross-discipline conflicts, review changes and surface evidence for a human reviewer.
The professional still decides whether the issue is valid.
That distinction matters because adding hundreds of unverified warnings would simply create another information problem.
The takeaway
The research does not support one magic rework percentage.
It supports something more important:
Bad project information, documentation errors, communication failures and unmanaged changes repeatedly contribute to avoidable cost.
The practical opportunity is therefore not to argue over whether rework is 1%, 3% or 10%.
It is to move discoverable problems upstream.
Find the contradiction before procurement.
Find the missing detail before fabrication.
Find the coordination issue before installation.
That is where document review becomes a financial-control activity, not just a design exercise.
Sources and further reading
- Autodesk / FMI — Harnessing the Data Advantage in Construction. Global survey of 3,900+ construction professionals; bad-data and rework estimates. https://investors.autodesk.com/news-releases/news-release-details/study-autodesk-and-fmi-finds-better-data-strategies-could-save
- Autodesk Construction — Data Advantage in Construction. Summary of project-data growth, poor decisions and contractor-level cost implications. https://construction.autodesk.com/resources/workflow-overview/harnessing-data-advantage-in-construction/