Find the step that stopped scaling before you cut the bill of materials.

Problem

Cost per node had stopped falling and nobody could say which part of the build was responsible. The instinct was to requote the bill of materials, because that is the number a spreadsheet shows.

Context

A four-person team building a soil-moisture node for smallholder farms. Runs of 100, 250, 250 and 400 units, all through the same contract assembler, over eleven months.

What they did

1. 1.Stopwatched every assembly step on run three and again on run four, at the bench, rather than reading the assembler's quoted line rate.
2. 2.Split the unit cost into parts, assembly labour and test, and tracked the three separately across all four runs.
3. 3.Requoted only the components whose share of the total had grown, which was two of thirty-one.
4. 4.Redesigned the connector interface so the hand-placement step became a single drop-in.

What to avoid

• •The cheapest bill of materials was not the cheapest unit. A regulator that saved 18 cents added a test step worth 41.
• •Quoted line rates flattered assembly by about a third. The bench numbers and the assembler's numbers disagreed for four runs before anybody checked.
• •Two months went into component requotes that moved the unit cost by under 2%.

Sources

• Run-by-run cost breakdownVerdant Sensing build log
• Bench timings, runs three and fourVerdant Sensing build log

Related lessons

• Keep 40% of the raise for batch two.→
• Budget for a second mould, not a perfect first one.→