The day your assembly house and your test house are two different vendors, you have added a coordination problem to the program — whether you planned for it or not.
Splitting assembly and test across best-of-breed suppliers looks efficient on paper. Each shop handles what it does best. Schedules get aggressive. And then the handoff arrives, and all the time you expected to save starts disappearing into the seam between the two vendors.
The boundary between an assembly shop and a test house is its own engineering problem. When it goes unmanaged, programs that looked clean on the Gantt slip for reasons that neither vendor can fully explain and neither vendor owns.
What lives at the handoff
Most of the schedule risk at a multi-vendor boundary comes from decisions that look like details but turn out to be load-bearing. Five things have to be settled before the first lot ships:
- A single owner for the data package. The traveler, the build records, and the as-built deltas need to cross the vendor line intact and in one place — not reconstructed from two half-copies after a failure.
- An agreed datum and handling specification. The test house needs to know exactly what it is receiving, how to handle it, and how to fixture it — before units arrive, not after the first lot is re-fixtured blind.
- Defined reject-disposition authority. Who decides scrap versus rework when a unit fails incoming? If that decision does not have a named owner, it defaults to a phone call that happens on the wrong day with the wrong information.
- A shared failure taxonomy. “Fail” has to mean the same thing on both sides of the boundary. Without a shared definition, a test-house reject and an assembly-floor re-inspection are measuring different things.
- Logistics and shipping eligibility resolved first. ITAR constraints, country-of-handling requirements, and re-export controls are not discovered at first lot. They are engineered in before the first shipment.
Why best-of-breed programs slip
The structural problem is that neither vendor owns the boundary. The assembly house owns what happens inside its walls. The test house owns what happens inside its walls. The space between them — the data package, the handoff protocol, the disposition authority — belongs to the program, which means it belongs to whoever is running the program.
When that coordination is explicit and documented from the start, the handoff becomes a transaction, not a negotiation. When it is not, each lot crossing the boundary is an opportunity to discover a new gap.
The cost of a bad seam
The irony of a split-vendor program is that the decision to use two specialized shops is usually correct. The assembly shop is better at assembly. The test house is better at test. The combination, on paper, beats a single shop doing both adequately.
But best-of-breed only beats single-source when the interface between the vendors is engineered as carefully as the processes inside each one. A program with two excellent shops and one bad seam is not better than a program with one good shop and no seam. It is slower and harder to debug.
The coordination work is not glamorous. It does not appear in either vendor’s capability sheet. But it is what determines whether the program lands on schedule or spends four weeks explaining to both vendors why the delay is not their fault.