CASES · PUBLIC VALIDATION
What we can show: measured, published and repeatable.
We do not ask for faith. We publish how we measure and what goes wrong, with parts anyone can download. Customer work is never shown without their written permission.
Measured result
98% accuracy on parts anyone can download.
- 423/431correct characteristics in STEP AP242
- 98.1%accuracy under the published criterion
- 0made-up values: doubtful items are flagged
- 0CAD licences needed
The case, step by step
How it was measured.
The challenge
Proving it without customer parts
To prove with data that the reading is reliable, without relying on customer parts we cannot show. For that we use the PMI test models published by NIST, the US National Institute of Standards and Technology.
The criterion
A match is a complete match
A characteristic only counts if the type and nominal match and, in feature control frames, the symbol, tolerance and datums. Each input format is measured separately.
The control
Doubtful items do not count
Anything the software cannot read with confidence is flagged “for review” and does not count as a match. Siemens NX 3D PDFs and scanned drawings are currently read with assisted review and improve with every version.
Why it matters
Anyone can repeat it
Any engineer can download the same parts and repeat the test. That is the difference between a marketing figure and a measured one.
Check it yourself
See the result before you install anything.
How we tell a case
Four rules we never break.
1
Always anonymous
We state the industry and the type of part, never the customer, the programme or the part number.
2
Only with permission
We never publish real work without the customer’s written permission.
3
Never a made-up dimension
Anything not read with complete confidence is flagged and confirmed by a person.
4
Reviewed by people
Every deliverable is reviewed by an engineer before it goes out.
More cases, with permission
Your part could be the next case.
We are starting our first pilots. When a customer authorises us in writing, we will tell their case here, anonymously or with their name.
Frequently asked questions
Why are there no named customer cases?
Because we do not publish real work without the customer’s written permission. In the meantime, we show what anyone can check: the validation with public NIST parts.
What happens with what GALGA does not read correctly?
It is flagged “for review”, with the reason, and a person confirms it. It does not count as a match in the validation.