Rework and scattered data, with their sources

Doing a job twice on site is a loss everyone knows. But it has no line of its own in the budget. It is spread across labour, material, crane hours and delay. This article looks at two questions through published studies: how much does rework cost, and how much of it comes from wrong or missing information.

The figures come from different definitions, countries and years. They cannot be added together. Next to each one is who measured it, how, and its limit.

Summary

The cost of rework as a share of project cost ranges from 0.38% to 12.4%, depending on the definition.
One survey attributes 52% of rework to poor project data and communication.
The newest study reports that project records understate the true cost of rework by about 300%.

The common finding is not a rate: a loss that is not recorded is not seen.

1. How much does rework cost

Studies of rework do not measure the same thing. One counts quality deviations, one counts field work, one counts rework from every cause.

12.4%: CII study published in 1987, 9 industrial projects. Quality deviations in design and construction as a share of total project cost. Limit: old and small sample; a wider definition than "rework".
4.5% (owner projects) and 2.5% (contractor projects): Hwang's 2006 study, 359 projects. Direct cost of rework from every cause, as a share of actual construction cost. Limit: general construction; not specific to piping and welding.
0.38% (before completion) and 0.76% (including after completion): Love's study of field rework, from ASCE's summary of 22 January 2026. Calculated from project records. Limit: the paper itself was not read, only the ASCE summary; the sample is not in the summary.

The same Love study reports that project records understate the true cost of rework by about 300%. So the lowest rate does not say "rework is low". It says "records are incomplete".

Evidence grade: measured (CII, Hwang, Love). For Love, only the ASCE summary was read.

2. How much of it comes from data

The 2018 "Construction Disconnected" survey by FMI and PlanGrid had about 600 respondents, mostly from US commercial construction.

52% of rework comes from poor project data and communication. The survey puts the annual cost of this to US construction at USD 31.3 billion.
Respondents reported spending more than 14 hours a week, 35% of their working time, on non-productive work: searching for project data (~5.5 h), resolving disputes (~5 h), fixing defective work (~4 h).
The survey puts the annual labour cost of this non-productive time to US construction at USD 177.5 billion.

Limit: survey-based, self-reported. Dollar figures are extrapolated from survey results. Not directly applicable to industrial plants.

The 2021 Autodesk and FMI report "Harnessing the Data Advantage in Construction" looked at a wider scale: a survey of more than 3,900 professionals, 2020 data.

The cost of "bad data", meaning inaccurate, incomplete, inaccessible or inconsistent data, to global construction in 2020 is estimated at USD 1.85 trillion.
The report puts USD 88.69 billion of this amount down to rework caused by bad data.
For a contractor with USD 1 billion annual revenue, the cost of bad data can reach up to USD 165 million.

Limit: a survey-based model estimate, not a measured loss. The contractor figure is an example scale calculation and an upper bound.

Evidence grade: industry estimate (two surveys).

3. What these figures do not say

They do not add up. CII measures quality deviations, Hwang rework from every cause, Love field rework. The FMI and Autodesk shares come from survey statements.
They are not specific to industrial piping. Most studies come from general or commercial construction. We found no separate rate for piping, welding and testing work.
They are not specific to Türkiye. We found no published rate for Türkiye.
They do not answer "how much is preventable". Saying a loss comes from data does not mean fixing the data ends the loss.
They are not measured losses. The dollar figures are estimates built from survey statements.

4. What it looks like on site

The figures are abstract. On site the same loss shows up in familiar ways:

Fit-up is done to an old revision of the isometric. The new revision is sitting in the engineer's email.
A weld fails NDE. Which welder made it, with which material batch, is written three different ways in three files.
The answer to "has this spool gone into a test package?" is in one person's head.
When a cut pipe is removed, why it was removed is not written down. The loss spreads into labour and material and never shows up as rework.

These examples are not measurements; they are the language of the site. Love's finding that records understate the loss lands in the same place as the last item.

5. What record it becomes in WeldTracker

This table does not say "this much less". It only shows in which record the loss becomes visible in the product.

Loss on site	Record where it is visible in WeldTracker
Rework, weld rejection	Weld log, NDE record, non-conformance record: which weld, who made it, which material batch
Searching for information	Work preparation (Workfront), document control (DCC), QR part passport
Revision disputes	Isometric revision tracking, red-mark board, conflict log
Rework that never gets recorded	Non-conformance record; removal and re-welding on the same weld line

"Common data environment" here is not a product label but a principle: the site crew and the office see the same record. AI reads isometric drawings and Excel files and makes suggestions. The engineer approves; what is approved goes into the record.

What we do not claim

We do not give a figure for how much WeldTracker reduces rework. When it is measured, we will publish it with the method.
The figures above are from industry studies; they are not WeldTracker's effect.
We did not use the shares in the Autodesk and FMI report about companies that use a common data environment and data standards. We have not verified them in the primary text.
We did not verify in the primary report the share (as a percentage) of bad data within rework, so we did not state it.
We have not yet found a rework rate specific to industrial piping or to Türkiye.

Sources

CII, Costs of Quality Deviations in Design and Construction: construction-institute.org
Hwang (2006), rework study of 359 projects
ASCE, "How much does field rework in construction actually cost?" (22 Jan 2026): asce.org
Love, Quantifying the Costs of Field Rework in Construction: doi.org
FMI/PlanGrid, Construction Disconnected (2018), Construction Dive summary: constructiondive.com
Autodesk/FMI, Harnessing the Data Advantage in Construction (2021): autodesk.com