Most inspection tooling is bought. Occasionally the thing you need does not exist, or exists in a form built for a different job, and the honest options are to work around it or to build it.
Smart Test Technologies is the part of the Test Inspect Group that builds it. This is what that means in practice, and — more usefully — how to tell when a bespoke tool is genuinely the right answer and when it is an expensive way to avoid a process fix.
Why an inspection group has a workshop at all
Because the people who know what is missing are the people doing the work, and that knowledge does not usually reach a manufacturer.
A technician who has spent a fortnight scanning the same awkward geometry knows exactly what fixture would halve the time and improve the repeatability. A Level 3 writing a procedure knows which calculation everyone gets wrong. An inspector filling in the same form for the eighth time that day knows which three fields are always the same and which one is always missed.
None of that is a product requirement anyone writes down. It reaches a workshop only if the workshop is in the same organisation as the work.
That is the whole rationale: STT exists to turn operational irritation into something reusable, and it draws its brief from the Group's own inspection, testing and training activity rather than from a market study.
Digital tools — the cheapest kind of engineering
The most-used things we have built are not hardware. They are calculators.
Welding and inspection are full of calculations that are individually trivial and collectively a reliable source of error: heat input from current, voltage and travel speed. Preheat from carbon equivalent, thickness and hydrogen level. Consumable consumption from joint geometry and deposition rate. Exposure from source activity, distance and thickness.
Every one of these gets done on a phone calculator, in a spreadsheet somebody built in 2019, or from memory. Each route produces occasional wrong answers, and the wrong answers are silent — a heat input recorded 15% low does not announce itself, it just sits in the record until an audit or a failure.
A web-based calculator with the formula fixed, the units labelled and the inputs validated removes that class of error entirely, and costs almost nothing to build once. Ours sit on the site, free, because the argument for charging for them was always weaker than the argument that everyone should be getting the same answer.
The quizzes work on the same logic. A technician revising for certification needs question banks with worked answers. Building them once and leaving them open costs us nothing and does something useful.
Reporting: the tool that changed the most
The largest thing built in-house was not an instrument. It was the way results get recorded.
Traditional inspection reporting has a shape that has not changed in thirty years: paper on site, transcription in an office, review, PDF, email. Each step is short and the queues between them are not, and the transcription step introduces errors that nobody catches because the person typing was not there.
Replacing it meant building structured, method-specific forms that a technician completes at the component on a tablet, with the acceptance criteria for that job already loaded, validation at the point of entry, and images attached where they were taken rather than emailed separately and reunited later by guesswork.
That is a software product, and it was built because no off-the-shelf system knew what a radiographic technique sheet or a welder qualification record needs to contain. Generic form builders get you 80% of the way and then require the last 20% to be a workaround, permanently.
Hardware and prototyping
Where the answer is physical, it is usually one of three things.
Fixtures and scanning aids — holding a probe at a repeatable angle and standoff on a geometry it was not designed for. Unglamorous, and often the difference between a technique that qualifies and one that does not, because repeatability is what a procedure depends on.
Handling and positioning equipment for the laboratory and the radiography compounds, where the constraint is getting a component into a defined position safely and the same way every time.
Control and interlock systems for facilities where access has to be governed — where a physical state, an authorisation and a safe condition all have to agree before anything happens. Safety-related work of that kind is designed to fail into a safe state rather than an operational one, which is a very different discipline from making something work.
The prototyping loop matters more than the manufacturing. Most first attempts are wrong in a way that only becomes obvious when somebody tries to use them in gloves, in the dark, at the top of a scaffold. Building three cheap iterations beats specifying one expensive one.
Building for people wearing gloves
Almost every design lesson we have learned the hard way reduces to the same thing: the conditions in the workshop are not the conditions on site.
Things that work at a bench and fail in the field: anything needing fine finger control, because gloves. Anything with a small screen, because sunlight. Anything that has to be held level, because scaffold. Anything with a fiddly connector, because it will be connected and disconnected forty times by someone in a hurry. Anything that needs a specific orientation to work, because it will be used upside down at some point and should fail obviously rather than quietly.
And anything relying on the operator remembering a sequence. If a tool can be used incorrectly, on a cold Tuesday at the end of a shift it will be, and the design question is whether that produces an obvious failure or a plausible wrong answer. Plausible wrong answers are the dangerous category — they get recorded, believed and acted on.
The practical consequence is that the useful prototype review is not "does it work", which it always does in the workshop. It is handing it to someone who does the job, with gloves on, and watching without helping.
When bespoke is the wrong answer
Worth as much as the rest of this article.
When something off the shelf does 90% of it. The last 10% is almost never worth designing, building, documenting and maintaining. A slightly awkward standard product beats a perfect bespoke one that one person understands.
When the real problem is process. A tool that makes a bad workflow faster entrenches the bad workflow. "We keep transcribing this wrong" is sometimes a form problem and sometimes a reason the data is being entered twice at all.
When it will be used once. Design and build time has to be recovered across uses. A single awkward job is usually cheaper solved with hire equipment and patience.
When nobody will own it. Bespoke equipment needs calibrating, maintaining, documenting and explaining to the next person. Without an owner it becomes an undocumented object in a corner that everyone is slightly afraid of.
We turn down more ideas than we build, and the ones we turn down are usually failing one of those four tests.
How something actually gets built
It starts with a job that went badly, or one that took three times longer than it should have. Somebody describes the problem rather than the solution — which is the important distinction, because "we need a jig" is a solution and "we cannot hold the probe steady on a 6-inch bend" is a problem, and the second has more answers than the first.
Then the boring questions: how often does this happen, what does the current workaround cost, and what does off-the-shelf actually do. Most ideas stop here, correctly.
What survives gets a rough prototype quickly and put in front of someone who does the job, with the explicit brief to find what is wrong with it. Then iterate, then document, then hand it over with someone's name against it.
If you have one of these
If there is a measurement you cannot take, a geometry you cannot get a probe onto, or a calculation your team keeps getting wrong, it is worth describing — including if the answer turns out to be that you do not need anything built.
Our free welding calculators and NDT quizzes are on the site now, and they came from exactly this route: somebody got tired of getting the same sum wrong.
TECHNICAL REVIEW — DELETE FROM THE LINE ABOVE, DOWN
Scheduled for 2026-09-10. Not to go live until a Level 3 or the RPA has read it.
Check: SMART TEST TECHNOLOGIES sub-brand piece, written to the /smart-test-technologies/ page positioning (product design and manufacturing function) rather than the homepage wording, which describes STT differently — see the flag in my report. The interlock/control paragraph is kept deliberately general; expand or cut depending on what is commercially sensitive.
Image to shoot: Workshop bench with a prototype fixture mid-iteration alongside the production version — the contrast tells the story better than either alone.
Internal links already in the text: /smart-test-technologies/, /weld-tools/, /digital-tools-inspection-technology/, /knowledge-base/ — confirm they read naturally, do not add more.
