Manufacturing: The Real Asset Is the Drawings, Not the Machines

A manufacturer's balance sheet lists the machines. It does not list the thirty years of drawings, tolerances, job records, material substitutions and hard-won fixes that tell you how to use them.

The machines can be bought by anybody with the capital. The rest cannot be bought at all, which makes it the only genuinely defensible asset in the business, and in most firms it is the least looked after.

This article is part of our guide to modernising a legacy business.

What the information asset consists of

Drawings and specifications, in whatever combination of paper, CAD files and superseded formats has accumulated. Usually organised by part number and findable only if you know the part number.

Tolerances and the reasons behind them. Which dimensions are genuinely critical and which were copied from a drawing in 1998 and have never been questioned. The difference between those two is a large amount of money.

Job records. What was quoted, what was run, what the scrap rate was, what went wrong. Frequently in a system at header level and in somebody's head at detail level.

Material and process substitutions. What you did when the specified material was unavailable, and whether it worked. Almost never recorded, and re-derived at cost every time it recurs.

The setup knowledge. How this machine behaves on this material. Which operator can hold that tolerance. What to do when the finish comes out wrong. This is the deepest layer and it exists entirely in people.

Why this matters now rather than eventually

Two reasons, and both are about people rather than technology.

The people holding it are retiring. This is a well-known feature of UK manufacturing and it is not slowing down. The setup knowledge and the substitution knowledge are held by the longest-serving staff, and neither is written anywhere.

Replacements learn it slowly and expensively. A new machinist becomes properly productive over a period measured in years, and most of what they are learning is the undocumented layer. Every retirement resets part of that clock.

The technology conversation follows from that, not the other way round.

What to do about it

Start with the substitution and fix log, because it is the cheapest and most immediately useful. A structured record of: this was specified, this was not available, we did this instead, it worked or it did not. Five fields, added to something people already complete at the end of a job.

Six months of that is more useful than a year of drawing digitisation, because it captures new knowledge rather than old, and because the people generating it are still available to explain it.

Then make the drawings findable by content, not just by part number. Being able to find every drawing that uses a superseded tolerance, or every part using a particular material, changes what you can do when a specification or a supplier changes. This is a document extraction project of the kind covered in scanned, filed and forgotten.

Then capture the setup knowledge from the people who have it, using the decision-recording approach in technology and succession. Record the reasoning out loud rather than asking anyone to write it, because they will not write it.

Then look at quoted versus actual by job type. Manufacturing is one of the sectors where this analysis most reliably finds something, because estimating conventions tend to be inherited and rarely revisited. See your pricing history is a dataset.

What to be sceptical about

The sector attracts a lot of talk about connected factories and predictive maintenance. Both are real, and both are frequently sold to businesses whose actual constraint is that nobody can find a drawing from 2011.

The test is the same as anywhere: what decision would you make differently? If the answer to sensor data is "we would know the machine is about to fail", ask how often that has cost you money, and compare it with what quoting the wrong price on a service line costs you every month.

For most UK manufacturers under two hundred people, the information asset is the bigger prize and the cheaper project.

Key Takeaways

• The machines are purchasable. The drawings, tolerances, job records and setup knowledge are not, which makes them the only defensible asset.
• The urgency is demographic. The deepest layer of knowledge sits with the longest-serving staff and is written nowhere.
• Start with a five-field substitution and fix log on jobs people already record. It captures new knowledge while the people who generated it are still there.
• Making drawings findable by content rather than part number changes what you can do when a specification or supplier changes.
• Be sceptical of connected-factory proposals if your actual constraint is that nobody can find a drawing from 2011.

Frequently Asked Questions

Our drawings are half paper, half CAD, across four formats. Where do we start?

With the ones still in active use, which is usually a much smaller set than the total. Digitising the full archive is rarely justified. Making the live set findable and consistent nearly always is, and it tells you whether the rest is worth doing.

How do we get long-serving staff to share setup knowledge?

By recording conversations rather than asking for documents, and by being clear that it is because the knowledge is valuable rather than because they are being replaced. Ask them to explain how they would teach it to someone starting Monday. That framing works considerably better than any documentation request.

Is AI useful in a manufacturing SME yet?

For the information problems above, genuinely yes: extracting from drawings, searching documentation, structuring job records. For anything touching machine control or safety, that is a different category with different rules, and it is not a blog-article decision.

---

Sitting on decades of drawings and know-how with nobody documenting it? Talk to Halo Technology Lab. Our AI solutions start with the information asset, and we will tell you when sensors are not your problem.