Why is it said that most errors originate in design?

Because up to 70% of the errors caught on site trace back to the design phase: incomplete information, late changes that never propagate, or data that stopped matching the model. And about 48% relates to communication and coordination failures between disciplines. The problem is almost never execution — it's the information that reaches it.

How do I measure rework if nobody logs it?

Start with three simple metrics: (1) hours spent redoing already-issued deliverables, (2) number of revisions per document before approval, (3) gaps found between budgeted and actual quantities. You don't need a system — a shared sheet for one month already gives you the scale of it.

Does automation actually reduce rework, or does it just speed up the work?

It reduces rework for a specific reason: it eliminates the manual step where the error gets introduced. When the quantity table generates by reading the model, it can't fall out of sync with the model. When standards checking is done by a tool, it doesn't depend on someone having a good day. Speed is the side effect; consistency is the main effect.

How long does it take to see the return on this kind of intervention?

Basic-level automations (generating tables, checking standards, exporting data) get built in 1-3 weeks, and the savings show up from the very first project review. The full return is usually visible within the first project cycle, because the rework avoided is immediate and measurable.

Does this apply to a company that doesn't work with BIM yet?

Partially. If the information lives in CAD and spreadsheets, there's a lot you can connect and automate without a full BIM rollout. But the biggest reduction in rework comes from having a single source of truth, and that does push toward the model. A process audit tells you what you can gain with what you already have.

Why doesn't the software we already pay for solve this?

Because the software covers the generic 80%. The remaining 20% is your company's own rules: how you measure, what format your client wants, what your quality control checks. No off-the-shelf tool ships with that, and it's exactly where the manual work that causes errors piles up.

← Back to blog

ReworkConstruction costsProductivityProject managementCivil engineeringQuality control

Rework on site: the 12-18% of the budget nobody measures (and how to cut it from the office)

Rework on site costs 12-18% of the budget and almost nobody measures it. Why it starts in design, and what you can fix from the office.

Zeist·August 6, 2026·Updated on October 8, 2026·10 min read

Share on LinkedIn

Hidden costs · Project management

Rework doesn't show up in any budget line item. But it gets paid for anyway.

Studies place the loss between 5% and 18% of project value. What stands out is where it comes from: up to 70% of the errors that blow up on site were already wrong back at the office.

The line item that isn't in the budget

Every construction budget has its line items: materials, labor, equipment, overhead, profit margin. None of them has a line that says "redo what went wrong."

And yet it gets paid. Every single month.

The available data is consistent, and uncomfortable:

Globally, between 5% and 10% of a project's total cost is lost to rework.
In Colombia, measurements have found up to 12.4% in rework alone, and between 12% and 18% when operational inefficiencies are added in.
Up to 70% of those errors originate in the design phase, not execution.
Roughly 48% relates to communication and coordination failures between disciplines.

On a million-dollar project, that's between 120,000 and 180,000 evaporating without anyone logging it as a loss.

Where the overrun actually starts

Design

70% of errors

Inconsistent information

Nobody catches it

Site

Discovered late

Rework

Paid in full

Where 12-18% of the budget goes

Rework doesn't show up in any line item, but it gets paid for anyway. Here's where it comes from, how to measure it, and how to cut it from the office.

01Diagnosis
The number nobody measures
Between 5% and 18% of the budget, depending on the study
02Root cause
70% starts in design
It's not the site that fails: it's the information reaching it
03Root cause
The 6 concrete origins
Stale data, mismatched criteria, crossed versions
04Measurement
How to measure it in your company
Three metrics you can start logging tomorrow
05Solution
What gets fixed from the office
The interventions with the best cost-to-impact ratio
06Action
Where to start without stopping production
A 90-day plan for a company already stretched thin

The first difficulty isn't reducing rework. It's seeing it.

Rework wears a disguise. It shows up as:

"The team's overtime this week"
"One more round of document revisions"
"A minor field adjustment"
"A change order the client approved"
"The delivery slipped a little"

None of those phrases sound like a loss. All of them are.

Why it goes unmeasured

There's no one to charge it to. If the error originated in design but was caught on site, whose cost is it? Since the answer is uncomfortable, it gets diluted.

It becomes normalized. After years, redoing deliverables is seen as "just how the business works," not as a fixable anomaly.

There's no box to check. Cost control systems track construction line items, not redone office hours.

This is the data point that changes the conversation.

When something goes wrong on site, the instinctive reaction is to look at execution: the contractor, supervision, logistics. But the studies point elsewhere: most errors were already baked into the information that reached the site.

What this looks like in practice

The drawing was out of date relative to the model.
The budgeted quantity didn't match the final design.
Two disciplines used different criteria for the same thing.
A routing change wasn't propagated to every document.
The construction detail didn't account for a clash visible in the model.

None of that gets fixed on site. It gets fixed earlier, at the office, where the cost of correcting it is a fraction of the alternative.

The 1-10-100 rule

A classic quality rule that applies perfectly here: fixing an error in design costs 1. Fixing it in planning costs 10. Fixing it on site costs 100.

It's not a metaphor: it's the difference between changing a number in a model and demolishing something already built.

In engineering and infrastructure projects, rework almost always comes from one of these six places.

1. The data lives in two places

The model says one thing and the spreadsheet says another. Someone exported it, adjusted it by hand, and from then on the two versions diverged. Every project revision widens the gap.

2. The criterion lives in someone's head

How a line item is measured, what's included, what's excluded. If only one person knows, the result changes every time someone else does the work. And when that person leaves, the criterion leaves with them.

3. Changes don't propagate

The alignment changes. The model gets updated, but not the quantities, not the budget, not the detail drawing. Every document left un-updated is rework waiting to happen.

4. Review depends on the human eye

Checking naming, layers, consistency across drawings. It's tedious work, and the error slips in precisely because it's tedious.

5. Every team delivers differently

Two sections of the same project, done by two people, come out with different formats and criteria. The time spent reconciling them is pure rework.

6. Information arrives late to whoever needs it

The site starts working off revision B while the office is already on revision D. Nobody did anything wrong — the workflow simply isn't connected.

The common pattern across all six

Manual step

The origin

Information splits off

Silently

Nobody catches it

Until it's too late

Rework

Gets paid

Notice the pattern: all six have a manual step where information can split off from its source. That's exactly where you intervene.

You don't need a management system or a consultant. You need three metrics and one month.

Metric 1 — Rework hours

Ask the team to log, in a shared sheet, the hours spent redoing something already delivered or already calculated. No detail needed: date, person, what was redone, how many hours.

What you'll discover: most companies are surprised by the total after the first month.

Metric 2 — Revisions per document

How many times the same deliverable gets issued before it's approved. If your documents routinely reach revision D or E, that's not rigor — something upstream is failing.

Practical benchmark: more than 3 revisions per document as a regular pattern signals a process problem, not client pickiness.

Metric 3 — Gap between budgeted and actual quantity

The gap between what the take-off said and what was actually built. It's the hardest-hitting metric, and the one that convinces leadership fastest.

How to present it

Rework hours per month        x  Loaded hourly cost   =  Monthly loss
Monthly loss                  x  12                   =  Annual loss
Annual loss / Revenue                                 =  % of the business

That last percentage is the one that unlocks budgets. If it comes out between 5% and 15%, you're in the industry's normal range — which is exactly the problem.

The interventions with the best cost-to-impact ratio, ordered by ease of implementation.

Level 1 — Connect the data to its source (weeks, high impact)

Have quantity tables, reports, and summaries generate directly from the model, instead of being exported and adjusted by hand.

What it eliminates: origin #1 and #3 from the previous module. If the table regenerates from the model, it can't fall out of date.

Typical cost: a tool like this is built in 1-3 weeks.

Level 2 — Automatic verification before delivery (weeks, high impact)

A button that checks naming, layers, required parameters, and consistency, and returns a report of what doesn't comply.

What it eliminates: origin #4 and a good chunk of #5. Review stops depending on someone having a good day.

Valuable side effect: the team learns the standard by using the tool, with no manual required.

Level 3 — Criteria written into the tool (weeks-months, structural impact)

Have the measurement rule embedded in the process, not in one person's head.

What it eliminates: origin #2, the most costly one long-term because it walks out the door when someone resigns.

Level 4 — Connected flow between office and site (months, high impact)

Have the published information always be the current version, with traceability of what changed.

What it eliminates: origin #6.

The plan has to work without stopping production. Nobody is going to halt projects to fix processes.

Days 1-30 — Measure without changing anything

Shared sheet to log rework.
At month's end, a one-hour meeting where the number goes on the table.
Identify the three tasks that generate the most rework.

Cost: zero. Just the discipline of logging.

Days 31-60 — Attack the first one

Take the number-one task on the list.
Write the exact criterion on one page: what's in, what's out, what rule, what exceptions.
Build or commission the automation for that single task.
Run it in parallel with the manual method for two revision cycles, comparing results.

Why in parallel: it builds team confidence and catches whether the criterion was misunderstood.

Days 61-90 — Consolidate and measure the effect

Replace the manual method with the tool.
Re-measure the rework metric for that task.
With the number in hand, decide whether to attack the second one.

By the end of 90 days, you have: a real measurement, one task solved, and a business case built on your own data to justify the next step. That's worth more than any theoretical presentation.

Where it costs more to fix an error, by stage

Aspect	Fix at the office	Fix on site
Relative cost	1x	100x
Time to fix	Minutes or hours	Days or weeks
Impact on schedule	Null or minimal	Chain-reaction delay
Impact on client relationship	None	High
People involved	1-2	The whole crew
Preventable with tools	High	Low
Logged as a loss	Almost never	Sometimes

The most common case of duplicated data: stop tracking your quantities in Excel.
How to get everyone delivering the same way: standardizing BIM processes in your company.
What exactly can be automated: automating Civil 3D, the complete guide.
If you're evaluating budget: how much an add-in costs.

Frequently asked questions about rework and hidden costs

Studies place the loss between 5% and 10% of total cost globally. In Latin America the figures run higher: in Colombia, measurements have found up to 12.4% in rework alone, and between 12% and 18% once operational inefficiencies are added. On a million-dollar project, that's between 120,000 and 180,000 that never shows up in any budget line item.

Did this resonate? Share it with whoever runs technical direction at your company. The first step costs nothing: measure for one month.

Rework doesn't show up in any budget line item. But it gets paid for anyway.

Where 12-18% of the budget goes

The number nobody measures

70% starts in design

The 6 concrete origins

How to measure it in your company

What gets fixed from the office

Where to start without stopping production

1. The number nobody measures

2. 70% starts in design, not on site

3. The 6 concrete origins of rework

4. How to measure it in your company (starting tomorrow)

5. What gets fixed from the office

6. A 90-day plan for a company that's already stretched thin

Frequently asked questions about rework and hidden costs

Keep reading on this topic

How to standardize BIM processes at your company without slowing production down (2026 guide)

BIM Audit: What It Is, What It Checks (Full Checklist) and When Your Firm Needs One

Automating Civil 3D: The Complete Guide (2026)

Is It Worth It for a Civil Engineer to Learn to Code? Real Benefits and What It Costs (2026)