How do you keep a cost engineer's know-how?

What goes missing when an experienced cost engineer leaves?

Take a twelve-person subcontractor where two people do the estimating. Their method lives in their head and in a spreadsheet only they can read well. When one of them leaves, the formulas stay and the reason behind every allowance goes.

For many shops that day is getting closer. According to Statistics Netherlands, 29 per cent of people working in Dutch industry in 2023 were aged 55 to 75, up from 19 per cent in 2013. In many firms the experienced cost engineer is also the manufacturing engineer who settles the route and the setups. That person takes two kinds of knowledge out of the door.

What goes is rarely a formula. It is judgement: which tolerance makes an extra setup risky, which material behaves differently from what the table promises, when a first-off part will run long. That shop knowledge sits in no ERP and no CAM project. Our research on the data gap comes to the same point: an integration can put the drawing, the order history and the NC program side by side, yet that does not by itself explain why an experienced manufacturing engineer picked one route and not the other.

Does an exit interview or a handbook help?

The usual response is an exit interview or a handbook of rules of thumb. The result is a document that falls short at the first unusual drawing. A rule of thumb without the part beside it is hard to apply and hard to check.

The second problem is timing. An experienced cost engineer knows a great deal, but only says it when a real part calls for it. ‘At that wall thickness I add a quarter of an hour’ does not come out of an interview. It comes out of the estimate for that one frame, at the moment of the decision.

Link Magazine, a Dutch trade title, described in 2025 how an ageing workforce is draining knowledge from manufacturing. Many firms have tried to standardise before, but delivering projects comes first and the agreements wear away. Knowledge retention works better when it is part of daily work than when it runs as a separate project.

For a shop with ISO 9001 certification this is also an audit point. Clause 7.1.6 of ISO 9001:2015 requires a company to determine the knowledge its processes need, maintain it and make it available to the extent necessary. An exit interview on its own rarely meets that. Its successor, ISO 9001:2026, was published in September 2026; according to UKAS, the UK accreditation body, certification bodies must move all their certified organisations to the new edition by 30 September 2029.

Record the know-how where it changes the price

An estimate is made of lines: material, setup, programming, machining and finishing. Each line has its inputs, such as a time, a rate and a quantity. That is the place to record know-how: on the line the experienced cost engineer adjusts, with the reason beside it.

The smallest unit for this is what we call a correction with reason. An example: ‘setup time from 60 to 90 minutes, second setup on a face with no datum’. That is one line of text. It sits with the part, the revision and the price it changed.

An ERP system does not do this. It keeps the order history and the price that came out, not the decision that led to it. A spreadsheet is worse: the cost engineer types 90 over the 60, and the old value and the reason are gone. On the repeat order a colleague copies the 90 without knowing why, or sets it back to 60 because the part looks simple. In that second case the half hour for the second setup is no longer in the quote, and the shop pays for it on every batch.

After a year, corrections with reasons add up to a correction history per part and per kind of decision: who changed what, when and why. That is know-how a colleague can check on their own.

What does the experienced cost engineer know, and where does it go?

The table sets kinds of know-how against the place you record them. It is a method, not software: it works on paper, in a spreadsheet and in an estimating package.

The pattern is always the same. Knowledge belongs with the object it is about: the line, the drawing requirement, the operation or the previous part. A separate folder of notes is never found at the moment it matters.

What do you do when everyone uses different cutting data?

At a seminar for machinists, a speaker asked the room to note the cutting data they would use to turn C45 down from roughly 60 to roughly 40. No two slips matched, the Dutch trade magazine Vraag en Aanbod reported in 2019.

The piece advises: agree, per material and per feature, on the tool, cutting speed, feed, width and depth of cut, and on how a rigid setup differs from a less stable one. Record the agreement centrally and adjust it when incoming material differs, rather than starting the search again at the machine.

For the estimate, the consequence is simple. As long as every machinist picks their own values, a cycle time in the quote has no fixed footing. With a shared agreement, a deviation in the actuals becomes a signal rather than noise. How to read that deviation operation by operation is covered in estimate and actual per operation.

Correction, measurement and calibration are three different things

At Subduxion we keep apart three things that are often blurred. A correction is a choice the cost engineer makes while estimating. A measurement is what actually happened in production. Calibration is the accepted comparison between the two, which shifts future default values.

That only works if every value carries its provenance: from a table, from the drawing, from a correction or from a measurement. An actual time is not representative by default either, because tool problems or other disruption end up in it too. Our research on production actuals separates representative variation, an explained exception and an unexplained anomaly.

So do not change a standard because one order overran. If several orders in a row overrun on the same milling operation, you correct the standard for that operation, not the whole estimate. That judgement stays with a person, which is what we mean by human-in-the-loop. We wrote earlier about how mature a time is in four cycle times for the same part.

Where do you start?

There is no need to wait for a big project. Four habits are enough to start tomorrow.

The test is simple: if a second person can follow the price without calling in the first cost engineer, the know-how is on record. Where a spreadsheet makes that test hard is the subject of Excel estimating.

1. Give every change to an estimate line a short reason in shop-floor language. ‘Adjusted’ or ‘override’ is not a reason.
2. Record cutting data per material and operation in one shared table, with the date of the last change.
3. For a repeat part, start with a repeat-part comparison against the previous estimate: what has changed, and why.
4. Have a second person check a few estimates each week, working from the reasons.