I have signed off on a dozen production transfers, and the pattern never changes. The machines arrive intact, the tooling is crated with care, the drawings travel in pristine condition — and within three months the receiving plant is scrapping parts the sending plant never scrapped. Nothing physical was lost. What vanished was twenty years of accumulated adjustments, workarounds and sensory judgement that lived in the heads of the operators, setters and maintenance technicians who ran the process daily.

Capability as measured on paper is a property of the process definition: parameters, tooling, specifications. Capability as it exists in reality is a property of the system that includes people. A transfer plan that relocates the hardware but not the human substrate produces a process that looks identical and behaves differently.

This article sets out how I structure requalification after a transfer, how I capture tacit knowledge before it walks out of the gate, and how dual-running exposes the gaps while there is still time to close them.

What Moves and What Stays Behind

Start with an honest inventory. When a process runs in one place for years, three layers build up. The first is documented: process sheets, setup instructions, maintenance schedules. This layer ships well. The second is parametric drift that was never documented — the press that needs a longer dwell than the standard calls for, the oven whose zone three runs hot and is compensated by a hand adjustment at loading. This layer ships only if you interrogate the current state aggressively.

The third layer is sensory: the operator who hears a bearing failing before the vibration sensor does, the setter who knows the tool needs changing by the look of the chip. This layer does not ship at all unless it is deliberately extracted and encoded before the last shift at the old site ends.

The failure mode I see most often is transferring the documented layer and assuming it represents the process. It represents the process as it was designed, perhaps a decade ago; the actual process has diverged. Before any equipment is disconnected, run a parameter reconciliation: compare the current set points in the controller against the master documentation and challenge every discrepancy. Half will be unauthorised drift that should be corrected; the other half will be undocumented corrections that saved the process and must be captured before departure.

Then walk the line with the operators and ask one question repeatedly: what do you do that is not written down? Ask the maintenance crew which faults never reach the CMMS because they are fixed in two minutes by habit. Ask the inspector which parts they pull aside on a feeling and what that feeling responds to. Film the setters doing a tool change. Record the sound of the machine in good condition. These artefacts cost almost nothing and are irrecoverable once the old site closes.

The three layers of a running process

  • Documented processProcess sheets, setup cards, maintenance schedules — travels with the equipment
  • Undocumented driftCompensating adjustments and corrected set points — recoverable only by parameter reconciliation
  • Sensory knowledgeSound, appearance and feel-based judgement — recoverable only by structured capture with people
Only the top layer ships by default; the two beneath it must be extracted deliberately or they are lost at decommissioning.

Requalifying the New Line as an Unknown Process

The worst assumption in a transfer is that requalification is a formality because the process is proven. It is proven in its old environment, with its old utilities, people and supply chain. The new site is a new process wearing the same clothes, and I qualification-plan it exactly as I would a brand-new line: a full process capability study against the same specifications, with sample sizes large enough to show the distribution has not shifted or widened.

The process that arrives here is not the process that left: capability must be re-earned on this floor, against these utilities, with these people.
The process that arrives here is not the process that left: capability must be re-earned on this floor, against these utilities, with these people.

Utilities deserve special attention. Mains power quality varies between sites; a voltage sag the old site never experienced can trip drives or corrupt weld schedules. Compressed air dewpoint, cooling water temperature and hardness, floor vibration from adjacent equipment, ambient humidity — each has bitten me at least once. On one aerospace transfer, a hard anodising line produced consistently different coating thicknesses because the new site's chilled water loop ran four degrees warmer, and nobody had listed water temperature as a process input.

Build a utility and environmental input matrix and verify every item at the receiving plant before the first qualification run. Treat the matrix as a controlled document: any input that differs from the sending site is a variable to be studied, not a footnote.

Requalification must also cover the measurement system. Gauges travel, but the calibration lab, the temperature-controlled inspection room and the fixture resting surfaces do not. A CMM that was levelled and foundation-mounted at the old site may sit on a mezzanine at the new one. Re-run gauge studies on the new floor before trusting any capability number coming off the line, and requalify the material flow: a different resin lot, a different steel mill, a different chemical supplier can shift the distribution even when every machine parameter is identical.

Capturing Tacit Knowledge Before It Evaporates

Tacit knowledge loss is the quiet killer of transfers because it never appears on the risk register as a line item. It manifests months later as unexplained variation, rising internal rejects and delivery misses nobody can diagnose. The people who could have diagnosed it in ten minutes are retired, reassigned or working for a competitor. The answer is structured knowledge capture — done early, done on camera, and done with the people who will never make the move.

Structured means you decompose the process into its decision points and ask at each one: who decides, based on what signal, using what criteria, and with what fallback? A setter deciding when to adjust compensation on a CNC grinder is judging on surface appearance, spark pattern and the last three parts' measurements. Write that decision logic down as if you were programming an apprentice's brain.

Where the criteria are genuinely sensory — the colour of a flame, the pitch of a sound — record the reference itself: a video of the good condition, an audio clip, a physical boundary sample retained in the standards cabinet. A photograph of a surface finish defect is worth a paragraph of text; the actual part is worth more than either.

Keep boundary samples of the good part, the marginal part and the reject at critical operations, agreed with the customer where relevant, and ship copies to the new site. Where a process relies on feel — torque by hand, fit by touch — the only transfer method is supervised practice, which leads directly to dual-running.

Dual-Running to Expose the Gap

Dual-running means operating the old and new lines in parallel for a defined period, producing the same parts from the same material lots, with outputs compared statistically rather than by eyeball. Its purpose is not extra capacity. Its purpose is to make the capability gap visible while the old process still exists as a reference standard and its people are still available to explain the differences. Once the old line is decommissioned, you have destroyed your control group.

Once the old line is decommissioned, you have destroyed your control group.

Structure the comparison properly: same input material, same tooling generation, interleaved production so time-based effects do not confound the result, and a defined protocol covering dimensional distributions, attribute defect rates by category, cycle times and first-pass yield. When the new line deviates, the diagnostic question is precise — what differs between the two environments? — and the old line becomes a living laboratory for testing hypotheses.

Move people as well as parts. Send the new site's setters to run the old line; send the old site's best setter to the new line for weeks, not days. Supervised practice at the new machine under the person who mastered the old one is the only reliable transfer method for judgement-based operations.

The exit criterion for dual-running must be explicit and agreed before the parallel period begins. My standard is consecutive production runs at the new site meeting the same capability and yield performance as the old site, with deviations resolved to root cause rather than patched with local adjustments. Cut dual-running short to save cost and you will repay it threefold in firefighting after the old line is gone. I would rather run parallel for an extra two months than explain to a customer why their part changed.

First Article, Then Volume: Qualifying in Steps

Jumping from installation straight to volume production is where transfers fail loudly. Stage the ramp deliberately. First articles off the new line get full layout inspection and, where the contract requires it, a fresh submission package — PPAP in automotive terms — treating the site change as a significant change, because that is what it is. Customers often have notification requirements for relocation; miss them and you risk a stop-ship on top of your capability problems.

Staged ramp-up after a transfer

  1. 01First articleFull layout inspection and customer submission treating relocation as a significant change
  2. 02Tightened early volumeHeavier inspection than the old line ever had, wound back only as data proves stability
  3. 03Defect-mix reviewCompare categories, not totals; a new mix points to an unidentified input difference
  4. 04Stepwise releaseVolume climbs in steps with a stability review at each gate
Each gate is a data decision, not a calendar date; volume only climbs when the previous step demonstrates stability.

Watch the defect categories, not just the total rate. If the new site produces the same overall reject rate but a different mix of defects, something about the process has genuinely changed, and the mix will tell you where to look. New-site-specific defects appearing at operations that were never problematic before are the fingerprint of an environmental or input difference you have not yet identified.

Suppliers to the line must be requalified in step. A local chemical supplier substituted for the original, chosen on price during the transfer project, is a classic source of delayed capability loss: everything runs well for six weeks, then plating adhesion fails. Freeze your supply base through qualification and early volume, and change one thing at a time. Transfers tempt managers to make several improvements simultaneously; resist, because you destroy your ability to attribute cause when something shifts.

Holding the Knowledge in Place Afterwards

The transfer ends when the new process has demonstrated sustained capability and the old line is gone — not before. At that point the job is to make sure the captured knowledge does not decay again. The videos, decision-logic documents, boundary samples and parameter reconciliations become the seed of the new site's process documentation, and they must be maintained as living documents: reviewed with operators, corrected when reality drifts.

Train the trainers. At least one person at the receiving site must own the transferred knowledge deeply enough to teach it and to notice when the process deviates for reasons the documentation does not explain — because it will, and the next undocumented correction starts accumulating immediately.

Across two decades of building and auditing quality systems in automotive and aerospace, I have seen that the plants which thrive after a transfer are the ones that institutionalised the habit of asking operators what is not written down — not just during the project, but permanently. That habit, more than any qualification document, is what keeps capability from quietly walking out of the gate the next time.