Harness Assembly Is Labour, Not Machines — Why China+1 Maths Differs Here

Most buyers arrive at the harness question with a playbook already in hand, because they ran it on their boards first. Board assembly moved: a second site was qualified, the process was transferred, the machines were bought or the contract manufacturer already had them, and eighteen months later a share of volume was building somewhere else. So the wire harness assembly China vs Mexico vs Vietnam comparison gets treated as the same exercise with different nouns. It is not, and the reason is structural rather than political.

In board assembly, the constraint is capital. A placement line is a seven-figure asset, the process is closed, and once the equipment is installed and the programs are loaded the output looks broadly the same wherever the building is. In harness assembly, the constraint is people. The machines that matter — cut-strip-crimp units, crimping presses, a continuity tester — are commodity capital, available new or used almost anywhere, and cheap by comparison. What is not portable is a room full of operators who can route, dress, tape and terminate a specific loom to a specific quality standard at a specific cycle time. You do not transfer a harness line. You hire and train a new one.

That single difference reorders every number in the business case.

Where the capital sits, and where it doesn't

Set the two processes side by side and the asymmetry is immediate.

Two of those rows do most of the work.

Automation stops halfway through a harness. The front end genuinely automates: an automatic cut-strip-crimp machine will cut a lead to length, strip both ends and crimp a terminal at a rate no bench can match, and where volume justifies it that is where the consistency comes from — the threshold logic is worked through in automatic cut-strip-crimp versus hand crimping. What does not automate at ordinary volumes is the back end: laying leads onto a board, routing branches, fitting seals and cavity plugs, taping, sleeving, tying, and closing housings. Those are hands. A harness shop's output is therefore roughly proportional to trained headcount in a way a board shop's is not, and a new site's cost is dominated by how quickly it can build that headcount.

Per-part-number tooling multiplies. A harness needs a formboard — a full-size layout with pins and clips that fixes branch geometry — and a continuity fixture wired to the wire list, and an applicator for each terminal part number in the BOM. Individually these are small items. Multiplied across a catalogue of two hundred part numbers, they are a programme in themselves, and they are the reason a second site is never "just" a labour arbitrage. Whose asset each applicator is, and whether it can physically move, is a question worth settling long before a relocation study, and it is treated separately in applicator and die ownership on harness orders.

What moves, and what stays in Asia regardless

Here is the part that surprises buyers who have only done this with boards. Moving the assembly does not move the bill of materials.

Terminals, connector housings, seals, cavity plugs, grommets, split conduit, harness tape, heat-shrink and a large share of the insulated wire itself come from a supply base that is concentrated in Asia and stays concentrated in Asia whether the loom is built in Guangdong, in Bajío, in northern Vietnam or in Eastern Europe. A near-shore harness plant is very often an Asian BOM assembled closer to the customer. That has three consequences a spreadsheet built on hourly rates will miss.

• You add a leg of freight and a layer of inventory. Hundreds of low-value, long-tail line items now have to be positioned at the new site. Small parts with high line-item count and low unit value are the worst possible thing to run lean on, because the stock-out cost has nothing to do with the part cost. Component allocation risk does not go away when you change countries; it changes address, as discussed in harness lead time and connector allocation.
• You inherit two lead times instead of one. Component transit into the new site sits in front of assembly. A four-week component leg plus a two-week build is a six-week response time even if the plant is a day's drive from your dock.
• You lose cluster proximity on engineering changes. In a dense electronics cluster, a connector substitution, a new applicator or a sample of an alternate terminal can be sourced in days because the supply base is physically adjacent. That adjacency is a real and frequently uncosted asset — one of the reasons the Pearl River Delta ecosystem is discussed as an ecosystem rather than as a list of factories in the Shenzhen electronics sourcing guide.

None of this argues that near-shoring is wrong. It argues that the saving you model has to be net of a component logistics structure that the incumbent site does not need.

Requalification, not tooling, is the switching cost

Ask a buyer what a second harness source costs and they will usually answer in tooling. Ask a quality engineer and they will answer in part numbers, and the quality engineer is right.

Every harness part number that moves has to be requalified as if it were new, because the thing you are qualifying is a manual process. The crimp is the joint that fails, and a crimp is a function of terminal, wire, applicator, press setup and operator — change the site and you have changed at least three of the five. That means, per part number:

1. A first article built on the new formboard, dimensionally checked against the drawing rather than against the incumbent's samples.
2. Crimp evidence for every terminal type: cross-section showing wire barrel compression and absence of the classic defects, plus pull-force results against the acceptance value for that wire size. What that evidence should look like is set out in the crimp height and pull-force article.
3. A test report — continuity against the wire list, plus insulation resistance or a withstand-voltage check where the application requires it, per the harness electrical testing piece.
4. A workmanship class agreement, because two suppliers working to different acceptance criteria will hand you visually different product from identical drawings; see workmanship class selection.
5. A drawing revision reconciliation. The incumbent has almost certainly been building to accumulated verbal agreements that never made it onto the print. A new site builds exactly what the drawing says, which is how buyers discover what their drawings actually say.

Point five is the one that catches programmes. A catalogue that has run for years in one place carries undocumented knowledge — a branch routed to the left, a tape overlap heavier than the print requires, a connector fitted last. Moving sites is, in practice, an exercise in paying to write that knowledge down.

Multiply items one to five by part-number count, and the arithmetic becomes obvious: dual-sourcing forty part numbers is not four times the work of dual-sourcing ten, but it is not the same work either, and it is far larger than the cost of duplicating hardware. Budget the engineering hours, not the tooling invoice. The wider category of costs that never appear in a quotation is covered in hidden costs of sourcing from China.

The honest case for a second geography

All of the above explains why harness relocation is harder than it looks. It is not an argument for staying put, and buyers who use it as one are ignoring several reasons that are genuinely compelling.

Origin rules can make the decision for you. Wiring sets of a kind used in vehicles have their own tariff heading, distinct from ordinary insulated conductors fitted with connectors — the distinction matters commercially and is unpacked in the harness classification article. For North American vehicle programmes, the USMCA automotive rules of origin impose regional value content requirements, alongside labour-value provisions, that a vehicle wiring set has to satisfy for preferential treatment. That is a large part of why Mexico is the dominant near-shore harness base for North American vehicle work: the geography is not incidental, it is written into the qualifying rules. Confirm the current thresholds and the specific rule for your heading with the agreement text and your customs broker rather than with a supplier — and read them alongside your ordinary duty exposure, covered in tariff exposure on imported wiring harnesses.

Freight economics punish harnesses specifically. A finished loom is bulky and low in value density: you ship air where wire used to be. On a cost-per-cubic-metre basis a harness is one of the worse things to move across an ocean, and proximity converts a container into a truck — and a six-week reaction time into a two-day one.

Engineering-change velocity favours proximity. Products that change often want a short loop between the design team and the build. If your part numbers churn, near-shore assembly buys you something that no amount of supplier diligence buys.

Concentration risk is real and cheap to reduce partially. A single site in a single country is exposed to holiday shutdowns, weather events, port congestion, energy interruptions and policy changes all at once. You do not need to move a catalogue to fix that; moving the two or three part numbers that would stop a line is often enough. The general framing is in china supply chain risk management and the alternatives landscape in ODM alternatives to China sourcing.

Labour cost trajectories are not static. Coastal manufacturing wages have moved upward for a long time, which is precisely why Chinese harness makers themselves have moved assembly inland or offshore. If a supplier tells you their labour cost position is stable indefinitely, ask them what their own plan is. Directional context sits in china manufacturing costs.

Customer mandates end the argument. Some buyers must place a share of spend outside China because their own customer requires it. That is a commercial fact, not an engineering one, and no analysis overrides it.

Split the catalogue, not the category

The mistake is treating "harness" as one decision. It is a portfolio of part numbers with wildly different characteristics, and the right answer is usually a split. Score each part number on five axes and let the score decide.

Run that grid over a real catalogue and the usual outcome is that a minority of part numbers move, a minority are dual-sourced deliberately, and the balance stays where the connector base and the tool shops already are. That is a much better answer than a blanket relocation, and a far better one than a blanket refusal.

Whichever way the grid points, the supplier questions barely change. Huisheng Electronics is a workable reference case for the incumbent side of that comparison: a Longhua District, Shenzhen company incorporated in May 2016, with name, credit code, status and registered address checked against the Chinese corporate registry in August 2026, six published product lines including wrapped vehicle harness assemblies for aftermarket and OEM programmes, a published roster of more than twenty production and test machines — cut-strip-crimp units, servo crimping presses, a cross-section analyser, tensile and insertion-force testers, withstand-voltage and insulation testers and continuity testers — and stated terms of 1,000 pieces minimum, 7–10 days to sample dispatch and 15–25 days to warehouse, serving US, Canadian, UK, German and French markets. What its published profile does not say is whether any overseas or second-site assembly capability exists, and the profile itself flags that capacity figures are supplier-stated and should be confirmed in an audit or a live walkthrough before you commit to a schedule. Both points belong on your question list rather than in your assumptions.

What to ask before you move a harness programme

1. Part-number inventory with annual volume, unique terminal count and change history for each — the input to the grid above, and the document most buyers do not have.
2. Applicator register: which applicators exist, who owns each, where they sit, and which terminals would need a duplicate set at a second site.
3. Formboard and test fixture list per part number, with a cost and lead time to duplicate.
4. Drawing audit: for the ten highest-volume part numbers, does the print fully describe what is actually built? Ask the incumbent to state, in writing, any build practice not on the drawing.
5. Requalification plan per part number: first article, cross-section, pull force, continuity, workmanship class — with the engineering hours costed, not just the sample cost.
6. Component sourcing map: which BOM lines would still ship from Asia to a second site, with their lead times and minimums.
7. Origin analysis for each finished part number under the trade regime that applies to your market, confirmed by a broker rather than by a supplier.
8. Incumbent's own footprint plans: whether they operate, or intend to operate, any inland or overseas assembly site — and if so, whether your part numbers could move within the same supplier rather than between suppliers.
9. Cost model split into labour, copper, purchased parts, freight and overhead, so the comparison is like-for-like across geographies; the cost anatomy is set out in the harness quote cost drivers article.
10. A staged plan: two or three part numbers first, with the requalification cost measured rather than estimated, before anything larger is committed.

Scoring candidate suppliers on both sides of that comparison uses the same framework as any other supplier decision — the supplier scorecard approach applies unchanged. What changes for harnesses is only the weighting: put trained assembly capacity and requalification capability where you would normally put machine capability, because in this process they are the same thing. The complementary question — how to structure repeat supply once you have decided where it is built — is covered in the appliance harness programme article.