Every plant you add keeps its own production record
How do we give auditors one traceability story across plants we acquired with different systems?
Figures and rules on this page apply to
United States
Working somewhere else? The shape of the problem usually travels. The deadlines do not.
What this looks like
A customer asks where a component lot went. One plant answers from its production system in minutes. The acquired plant pulls paper travellers from a filing room. The third exports a spreadsheet with different column names. Quality stitches the answer together by hand while the customer waits.
The numbers
Every figure here is someone else’s. Check them.
- 2 Feb 2026date the FDA's Quality Management System Regulation took effect, bringing in ISO 13485:2016 by reference
- 35%share of FY2020 device inspection findings for corrective action, complaints and design controls combined
- 197corrective and preventive action citations issued to device makers in FY2020, the most cited area
Why it happens
It is not a people problem.
Each plant records production correctly, in its own vocabulary. One calls it a lot, another a batch, another a work order. Their fields line up on paper and mean different things. Traceability breaks where the vocabularies meet, not inside any one plant.
Why your current software has not fixed it
Because it was never built to.
The standard answer is one quality system across every plant. That is a multi-year validation project, and the next acquisition closes before it finishes. Until then, the gap between plants is covered by the people who remember how each site numbers things.
Intelligence, plumbed in
A scanned traveller with handwritten lot numbers is readable, and it can be tied to the right batch.
We build the layer that says what each plant's record means, then read every plant's records through it. Underneath it: a connector, one agreed meaning per field, and a test set scored on your own records.
How we make AI survive real data-
Connectors -
A semantic layer -
Evals you can check
However hard, whatever it is
Acquired plants are one example. Bring the record nobody can find in time.
We wrote this for contract device makers because the new regulation makes it current. The same approach works for any regulated maker running more than one site.
-
01 We sit with you
Days where the work happens, not a workshop in a meeting room. We watch the job get done and write down the shortcuts nobody wrote down.
-
02 We read everything
Your data, your rules, your vendors and their documentation, and the published research on your sector. We report what is actually in there.
-
03 We break it to first principles
Not which tool fixes this. What is actually causing it, taken apart until we reach the piece that cannot be divided further.
-
04 Then we build
Weeks, not quarters. By this point we are not guessing what to build, and guessing is the thing that makes projects long.
What we build
Specific enough to argue with.
Four mechanisms, not four features. Each one is a thing that happens on its own, every day, whether or not anyone remembers to run it.
-
Travellers, inspection sheets and certificates read from every plant, paper and scanned included.
-
One agreed meaning for lot, batch and work order across sites, held above each plant's own system.
-
A traceability answer, component to finished device, put together in minutes for any lot at any plant.
-
Each newly acquired plant connected to the same layer in weeks, not after a validation cycle.
How you would know it worked
Numbers in your own reporting, not ours.
- Time to answer a customer or auditor traceability request, per plant.
- Share of production records captured without manual re-entry at acquired sites.
- Traceability findings at the next audit.
Straight answers
Where a model is involved, it is scored against your own records first. Accuracy per source, not one flattering average.
The questions this raises
-
We plan to move every plant onto one quality system. Why do this first?
The move needs the same mapping between plants anyway. Building it first gives auditors one answer now and shortens the migration.
-
Does this need to be validated?
Anything that feeds a quality decision is validated for its intended use. We scope it so the record of truth stays in your validated systems.
-
What about plants that still run on paper?
They are the usual case. Scanned travellers and handwritten entries are read and linked like any other record.
Where the figures come from
We did not make these up, and you should not take our word for them.
Also in manufacturing
Productivity fell in fourteen of the twenty manufacturing groups measured
Federal figures show labour productivity declining across most manufacturing industry groups. Machine data is abundant. The work between machines is not recorded at all.
ProblemEvery hotel project arrives as a different drawing set
Contract furniture makers win hotel and retail projects on bespoke work. Each brief arrives as someone else's drawings, schedules and specs, and quoting waits for a person to read them all.
ProblemEvery line you built is one of a kind, and its history is in service notes
Makers of packing and sorting lines build each one to order, then service it for years. The real record of each line sits in service reports and engineers' notes.
ProblemEach shop in the group quotes alone, from its own past jobs
Groups of contract manufacturers grow by buying local shops. Each shop still quotes from its own records, so the group never sees what it already knows.
ProblemThe shop you just bought can make parts your estimators never quote
Precision machining groups grow by buying shops. Each shop arrives with its own machines, job history and quoting habits, and requests still land wherever the customer first sent them.
ProblemOne heat network, two pipe ranges, and every bid costed by hand
Pipe makers now sell complete heat-network systems. Each bid still means reading someone else's network plans and costing every run by hand, often in two separate catalogues.
The meaning is not in any schema
Two systems agree on the field and disagree on what it means.
The rule lives in a document, or in what one site has always called something, or in a habit nobody wrote down. Joining the data is the easy half; agreeing what it means is the half that is actually the work.
- Behavioral healthOne in four facilities still run on paper, and there is a reason
- Accounting & consultingAn eight-hundred-billion-dollar quarter built on hours recalled from memory
- Senior livingNobody wants a sixth dashboard
- Skilled nursingEvery building reports. Nobody can see the portfolio.
- Supply chainThe distributor you just bought still takes orders its own way
Consent rules, site vocabulary, what an hour counts as. Wherever two systems agree on a field and not on what it means.Describe the disagreement.
Traceability across plants is a translation problem, not a replacement project.
We build the layer that says what each plant's record means, then read every plant's records through it.
See everything we build-
Software -
Hardware -
Ways of working -
Whole ventures
Is this happening to you? Tell us the size of it.
Twenty minutes. We will tell you honestly whether the numbers justify doing anything about it.