Networking hardware · Supply chain
Qualify the second supplier before you need one.
Run one paid pilot batch a year with a supplier you are not using.
Switched assemblers with a seven-week gap in shipments
Adaeze Okoro
@adaeze-networks
Written by someone who worked on it.
A contract assembler went from responsive to unreachable in three weeks. The switch took four months, and three of them were qualification the team could have done a year earlier.
Problem
The only assembler who had ever built the product stopped answering. There was no second source, no transferable test fixture, and no documentation that was not in that assembler's head.
Context
A router for community networks, built in batches of 500. Single-sourced from the start because the first assembler had been generous with a small team.
What they did
- Rebuilt the test fixture from the schematic rather than asking for the original back.
- Wrote the assembly instructions from a filmed build rather than from memory.
- Paid a second assembler for a 50-unit pilot before committing a production batch.
- Kept the pilot going at 50 units a quarter after the switch, so a third source was never a cold start.
What to avoid
- The test fixture turned out to encode three undocumented calibration constants, found only when the new assembler's yields came back wrong.
- The first assembler's quoted price had been below cost for two years, which nobody discovered until they got a market quote.
- Qualification was scoped as a purchasing task and took an engineer full-time for six weeks.
The business facts
- Timeline
- 4 months to switch, 7 weeks with no shipments
- Mesh & Mortar
- Lagos, 2023
- Weeks without shipments
- 7
- Unit cost change after the switch
- 22.0%
- Pilot units before committing
- 50
Sources
- Assembler transition post-mortemMesh & Mortar engineering notes
Related lessons
- supply-chain
- contract-manufacturing
- risk
44,190 views