Outsourcing Structural Engineering: What Actually Goes Wrong

The package comes back on the day it was promised. The model runs. Every member passes. The drawings are tidy, the calculation sheet is legible, and the person who did the work clearly knew what they were doing.

And your engineer still spends four days fixing it.

This is the version of outsourcing that nobody warns you about, because it doesn't look like failure. There is no obvious incompetence to point at. There is just a package that was built on a slightly different set of assumptions than the ones in your office, and now somebody has to reconcile the two — usually the most expensive person on your team, usually the week the drawings were due.

We sit on the production side of this arrangement. We prepare structural analysis, design and documentation for firms in other countries, and their licensed engineers review and seal it. So we've watched a fair number of these engagements start well, and we've watched some of them go sideways in the first month. The pattern is consistent enough to be worth writing down.

Almost none of it is about engineering ability. It's about handover.

First, the reason this keeps happening

Small technical firms don't have a smooth pipeline. They have a lumpy one. Three tenders land in the same fortnight, then six quiet weeks, then a developer wakes up and wants the whole podium re-run.

A full-time hire is a flat line drawn across that. You're paying the same salary in the quiet weeks as in the brutal ones, and if the brutal weeks are only a third of the year, the maths never quite closes. So the work goes out — to a freelancer found on a platform, to a friend-of-a-friend's office overseas, to whoever answered the email.

That decision is usually sound. What follows it usually isn't, because the brief that goes out with the work is written in about eleven minutes by someone who is already behind.

Failure 1 — The code was named. The edition wasn't.

"Design to ASCE 7" is not a design criterion. It's a family of documents that disagree with each other.

ASCE 7-22 revised the basic wind speed maps, with most of the change concentrated in hurricane-prone regions, and added an entire new chapter on tornado loads — Chapter 32 — which requires Risk Category III and IV structures inside the mapped tornado-prone region to be designed for the greater of tornado or wind demands. That standard is referenced by the 2024 IBC. If your jurisdiction is still on an earlier IBC cycle referencing ASCE 7-16, and the offshore team designs to 7-22 because it's newer and therefore assumed better, you now have a package that doesn't match the code your plan reviewer is holding.

The same trap exists everywhere. AS/NZS 1170.2:2021 superseded the 2011 edition and its amendments, and it didn't just adjust numbers — it redrew wind regions. Central Australia moved into a new Region A0. Tasmania became Region A4. Region B split into B1 and B2. A climate change multiplier came in. There are dozens of substantive changes in that revision, and a site classified under the 2011 map can land in a different region under the 2021 one.

So when a package is prepared to the wrong edition, the failure isn't a rounding error. It's a different wind region, a different load, and a redesign.

Failure 2 — Nobody wrote down who checks what

Every firm we work with has a licensed engineer who reviews and seals. That part is never in dispute. What gets fuzzy is everything upstream of the seal.

There is a real difference between "check that this is right" and "check that this matches what we would have done." The first is a technical review. The second is a house-style review, and it takes about three times as long, because it isn't looking for errors — it's looking for divergence.

When that expectation isn't stated, the reviewing engineer opens a technically correct package, finds that the shear wall was placed where the analysis wanted it rather than where the office always puts it, and treats that as a defect. It isn't a defect. It's an unbriefed preference. But it burns the same hours.

Write down three things before the first package: who reviews, what they're reviewing against, and how many review cycles are inside the fee. That last one matters more than people expect — an engagement with unlimited revisions built into a fixed price is either mispriced or about to become adversarial.

Failure 3 — The quantities came off a different revision than the drawings

This one is quiet, and it's expensive, and it's the reason we don't let takeoff sit in a separate silo from documentation.

Here's the mechanism. Drawings go to Rev C. A column drops from 500 square to 450 in one grid line. The BOQ was extracted at Rev B, because that's when the quantity surveyor had a free afternoon. Nobody re-extracts, because the change looks small and everyone remembers it. The tender goes out with a concrete volume that doesn't match the drawing set it was priced from.

Nothing catches this. Not a structural check — the design is fine. Not a drawing review — the drawings are current. The two documents are individually correct and jointly wrong, and the discrepancy surfaces on site, when a subcontractor prices a variation.

The discipline that prevents it is unglamorous: quantities are re-derived from the issued revision, every issue, and the reconciliation is recorded. When a BOM or BOQ leaves our office it carries the drawing revision it was taken off, on the face of the document. It's a small thing. It has saved a lot of arguments.

Failure 4 — Your office standards were assumed, not sent

Firms underestimate how much of their drawing output is convention rather than engineering. Layer naming. Title block fields and who fills them. Which details come from the standard library and which get drawn fresh. Whether bar schedules go on the sheet or in a separate table. Whether sections are cut looking north or looking at the grid.

None of that is in the code. All of it is in your reviewer's head, and none of it travelled with the brief.

So the first package comes back drafted to somebody else's convention, and it reads as sloppy even though every line is deliberate. The reviewer's confidence drops on the first sheet, and once confidence drops, everything after that gets read as a mistake.

The fix costs one email: send a completed drawing set from a project you were happy with, plus your CAD template. A competent offshore team will match it on the second package and hold the match after that. If they can't match a template you handed them, you've learned something cheaply.

Failure 5 — The feedback happened on a call

Time zones get blamed for a lot of outsourcing problems they don't actually cause. Asynchronous work is fine. What isn't fine is verbal feedback that never becomes a written record.

A call is comfortable. You talk through eleven changes in twenty minutes and everyone hangs up feeling aligned. Then nine of the eleven get made, one gets half-made because the phrasing was ambiguous, and one gets forgotten entirely. Nobody can prove what was agreed, so the argument becomes about memory rather than about the drawing.

Marked-up PDFs solve this completely. The comment is attached to the location it refers to, it survives the week, and it can be closed out item by item. Use the call for the things calls are good at — judgment, priorities, why a decision was made — and put every actionable change in writing on the drawing itself.

The scoping sheet

Everything above collapses into a handful of questions answered before work starts. This is roughly what we ask for, and you should expect any serious partner to ask for something similar:

Eight lines. Fifteen minutes to fill in. It is genuinely the difference between an engagement that works and one that quietly costs you more than the hire you were avoiding.

What outsourcing will not fix

We would rather say this plainly than have you find out on a live job.

Outsourcing does not fix an unclear design intent. If the architectural scheme is still moving, sending it offshore just means the churn happens at a distance, with a lag, and with a variation attached to each lap. Settle the scheme first.

It does not remove review time. It moves the work, not the responsibility. Whoever seals the drawings has to read them properly, and that engineer's hours are a real cost that belongs in your fee calculation from day one. Any comparison of hourly rates that ignores review time is a comparison you'll regret.

It does not replace local knowledge. A jurisdiction with an unusual plan-review culture, a specific local amendment, or an inspector with known preferences needs someone local who has been through it. We can design to the standard. We can't tell you what the reviewer in your county is going to ask for in the third week.

And it doesn't survive a bad first brief. Which is the whole point of this article.

How to start without betting a quarter on it

Don't begin with a retainer. Don't begin with the biggest job in the office, either — the instinct to "make it worth the setup effort" is exactly backwards.

Begin with one small, complete, real package. A single-storey steel frame. One foundation package. An existing-structure check. Something with a genuine deliverable at the end, small enough that a bad outcome costs a week rather than a project.

That one package tells you almost everything: whether they handled the code edition without being reminded, whether the drawings look like yours, whether the quantities tie to the revision, whether the questions they asked mid-way were good questions, and whether the reply came when they said it would.

If the answers are yes, the second package is easy and the third one barely needs supervision. If they're no, you've spent a week finding out instead of a quarter.

Start With One Package →