The pilot run: building under production conditions before you depend on it

Decide what the run is testing before scheduling it

Pilots drift into vagueness when the objective is simply to build some units. Write the questions the run must answer: does the tooling produce conforming parts across a genuine cycle, can operators follow the instructions as written, does the test coverage catch the failures that occur, does the assembly sequence work at rate, does the packaging survive handling, is the paperwork produced correctly without an engineer standing over it. Each question implies something to observe and someone to observe it. Anything not on that list can be checked later; anything on it deserves a person watching rather than a record read afterwards.

Production conditions means production people

The most common way a pilot flatters a process is by staffing it with engineers. Your specialists and the manufacturer's technical team know the product intimately, work carefully, and unconsciously correct problems that a line operator would either escalate or miss. Let regular operators build it, on their normal shift, using the released work instructions and the training they were actually given. Engineers should be present and watching, not participating. What they see is the point of the exercise: where somebody hesitates, improvises a fixture, asks a neighbour, or does something the instruction never described.

Measure the process, not only the parts

Conforming units are a weak result on their own, because a determined team can produce good units from a poor process. Record what it took: cycle time per station and where the line waited, scrap and rework with causes attached, how many units needed a second pass at test, setup and changeover duration, material consumed against material planned, and every stoppage with its reason. That data tells you whether the quoted cost is realistic and whether the promised rate exists. It is also the baseline against which later drift becomes visible, and it is only collectable while somebody is deliberately watching.

Treat the output as evidence rather than inventory

Pilot units are built under abnormal supervision, sometimes with temporary fixtures, occasionally with material that was expedited. Selling them is tempting when the run goes well and demand is waiting, and it removes both your ability to destructively test and your freedom to conclude that the build was unsatisfactory. Decide the disposition in advance: what will be torn down, what goes to environmental or life testing, what is retained as a reference for future disputes, and what may be sold if and only if defined criteria are met. Where a regulated product is involved, the release rules govern that decision, not the schedule.

Close the findings before the next build, with names against them

A pilot generates a list, and the list is worthless without ownership and a deadline that sits before the following build. Split it into items that must be fixed and reverified, items that can be tracked into early production, and items accepted deliberately with a reason recorded. Some findings belong to the manufacturer, some to your design team, and some to whoever wrote the documentation, so assign them honestly rather than defaulting everything to the supplier. Then agree what evidence closes each one, because a pilot whose actions are closed by assertion has taught you nothing you can rely on.

Frequently asked questions

How large should a pilot run be?

Large enough to exercise the things you are trying to observe. That usually means more than one setup, at least one changeover, a full material batch where batch effects matter, and enough units for any statistical check you intend to run. A build too small to show variation only proves the process can be nursed. Scale also depends on the cost of the units and how destructive your testing will be, so set the quantity from the questions rather than from a habit.

Who should attend the pilot, and does the customer need to be there?

Someone from your engineering team and someone from quality should be present for the whole build, not for a scheduled visit within it. The manufacturer's process engineer, the line supervisor and their quality engineer should be available throughout. Bringing your own major customer is occasionally valuable where they hold approval rights, and often unhelpful, since a pilot is meant to surface problems and an audience changes how openly people report them.

What if the pilot goes badly — do we repeat it?

Repeat it when the failures were systemic, meaning the process, documentation or training could not support the build. Proceed with tracked actions when the issues were specific, understood and fixed with evidence. The judgement to avoid is treating a bad pilot as bad luck and running the same build again hoping for a better day. Record which findings were closed and which were carried, because a launch that then struggles will be investigated against exactly that decision.