Module 5 · Lesson 5.1
The chain of handovers
What a workflow actually is, what each step adds, and the five things nothing carries.
Why this matters
Ask an engineer to describe their workflow and you usually get a list of software. That description is useless for finding where things go wrong, because the software is not where things go wrong — the gaps between the software are.
A workflow is better described as a chain of handovers. At each one, information is added, information is lost, and responsibility moves. Describe it that way and the failure points become obvious, because they are all in the same place: the arrows, not the boxes.
By the end of this lesson you should be able to
- Describe a workflow as handovers rather than as tools
- Say what each handover adds that did not exist upstream
- Identify who owns a handover, as distinct from who performs it
- Name the attributes that survive no handover in the chain
What you should already know
- Module 3's parametric model — the second box in the chain
- Module 4's scripting — most handovers are automated by somebody
Boxes and arrows
A typical structural workflow has five handovers:
| From | To | What is ADDED here |
|---|---|---|
| Architectural model | Parametric model | Grid, levels, and the decision about what is structure |
| Parametric model | Analysis model | Restraints, releases, sections, materials, loads — everything that makes it behave |
| Analysis model | Member design | Effective lengths, combinations, and the code the check is made to |
| Member design | Coordination model | The sizes everyone downstream will build to |
| Coordination model | Fabrication model | Connections, holes, welds, cuts |
Read the right-hand column again. Every handover adds something that did not exist upstream, and the thing it adds is engineering judgement. The arrow is not a pipe; it is a person making decisions.
Which is why 'the model' is a phrase worth distrusting. There are five models here, none of them is a version of the others, and no one of them is authoritative about anything the others contain.
Who owns a handover
Ownership is not the same as performance. A graduate may run the export; the person who owns it is whoever is answerable if the receiving model is wrong.
The useful test: who would be asked to explain it? If the answer is 'nobody in particular', that handover has no owner, and it is where the next problem will come from.
What survives
Run the five handovers against the twelve things a structural model carries and one result stands out. Five attributes survive no handover in the chain:
- Supports and restraints
- End releases
- Load cases
- Load combinations
- Analysis results
Every one of them has to be re-created, by a person, at every stage. That is not a defect in any particular format — it follows from what the formats are for. A geometry interchange has no concept of a support because it has no concept of a structure. A fabrication interchange knows a great deal about a member and nothing about the frame.
The last one on the list is the important one, and it is the one rule in this module that has no exceptions.
Never exchange analysis results. Re-run them.
A received force is a number with no model behind it. It cannot be checked, it cannot be traced to the geometry it came from, and it will still be there long after the model that produced it has changed.
Formats carry what they were built for
It is worth being precise about why a format loses something, because the reason predicts what else it will lose.
- A geometry interchange carries shape. It has no member concept at all, so everything structural goes, and the picture still looks right — the combination that causes the most trouble.
- An object model interchange carries what a member is, for coordination. It carries much less about how the member behaves.
- A steelwork interchange is excellent at the member and silent about the structure. It was never meant to carry an analysis model.
- A coordinate table carries whatever the person building it decided to include, and the gaps are invisible because nobody wrote down what was left out.
None of these is a criticism. A format that carried everything would be a native file, and a native file is only lossless while the version is the same — which is a narrower guarantee than it sounds.
Try it
Digital workflow builder
One handover at a time. What arrives, what does not, and who owns the gap.
1 of 5
Before revealing, decide for yourself which of the twelve attributes this route carries.
Architectural model → Parametric model
- Exchange route
- Geometry interchange (lines and points)
- Owned by
- Structural lead
- Adds
- Grid, levels, and the decision about what is structure.
Twelve attributes go into this handover. Decide which arrive before revealing.
Across the whole chain
- Attributes surviving no handover
- 5
These are re-created by a person every time
- Supports and restraints
- End releases
- Load cases
- Load combinations
- Analysis results
What this shows: Every handover adds engineering judgement, loses information, and belongs to somebody — and five attributes survive none of them.
Worked example
What a coordinate table actually carries
Given
- A truss is passed to a checking engineer as a spreadsheet of node coordinates and member ends
- The sender's model had pinned supports at one end, a roller at the other, and 4000 mm² chords
- The receiving engineer builds an analysis model from the table
Find
Which of the twelve model attributes reached the receiving engineer
Practice
A workflow has five handovers. If each one is performed correctly 95 % of the time, and the errors are independent, what percentage of the time does a model reach the end of the chain with no handover error at all?
Practice
Of the twelve attributes a structural model carries, how many survive no handover at all in the five-step chain?
Check yourself
An analysis model is exported to a coordination model, and the end releases are lost. Is this a problem?
Worked example
Who owns the handover nobody set up
Given
- A parametric model is exported to analysis by a graduate, using a script written by a colleague who has left
- The script was never documented and has been used on four projects
Find
Who owns this handover
Practice
A chain of four handovers is each performed correctly 98 % of the time. What percentage of models reach the end with no handover error?
Practice
Of the twelve model attributes, a coordinate table carries two intact and loses six outright. How many does it degrade?
Check yourself
Why is 'the model' a phrase worth distrusting on a project?
Check yourself
What does every handover in a structural workflow add?
Check yourself
Why should analysis results never be exchanged?
Check yourself
A format carries geometry perfectly and no structural properties. What makes that combination dangerous rather than merely limited?
Summary
- A workflow is a chain of handovers, not a list of software
- Every handover adds engineering judgement that did not exist upstream
- Ownership is who would be asked to explain it, not who ran the export
- Restraints, releases, loads, combinations and results survive nothing
- Never exchange results — a received force has no model behind it
- The right question is what the receiving model needs to do
This is educational material. It uses simplified examples to teach principles, and must not be relied on for real design or safety-critical decisions. Module overview and checkpoint