Pilot production: proving a process before the plant is committed

The output is data; product is incidental

Committing space and people to a pilot means agreeing that the line will be run inefficiently on purpose. Batches are deliberately varied to map operating windows, failures are provoked rather than avoided, and changeovers happen far more often than any production schedule would tolerate. That is the arrangement's value: knowledge bought at small scale costs a fraction of knowledge bought after commissioning a full line. It also means the pilot cannot be judged on utilisation or unit cost, and a plant manager measured on those numbers will quietly starve it.

Which products justify the detour

Not everything needs a pilot. Processes worth piloting are those whose behaviour changes with scale or where being wrong is severe: chemistries sensitive to mixing and heat transfer, coatings, fermentation, novel joining methods, anything requiring qualification batches before approval. Products assembled from characterised parts on familiar equipment usually do not need one, since a planned trial run on the production line answers the question. Demand shape matters too — a launch into an uncertain market benefits from a stage where volume can be supplied without committing the full asset. Piloting something that needs no pilot merely delays the launch and consumes engineering time that was scarce anyway.

Small equipment that has to predict big equipment

The hard engineering problem is representativeness. Geometry, agitation, surface-to-volume ratio, residence time and heat removal do not scale proportionally, so a pilot vessel that mixes beautifully can flatter a process destined to struggle in a production reactor. Capital goes into instrumentation rather than capacity: more sensors, sampling points, data logging and controllable variables than the eventual line will ever carry. Where possible the pilot should use production-standard components in reduced numbers — one filling head instead of many, an identical extruder — because similarity of mechanism predicts far better than similarity of size.

Materials, suppliers and the cost of learning

Pilot quantities attract poor prices and awkward minimums. Suppliers send development samples that may not represent their routine output, a common source of surprises when the plant scales and the material shifts within its own specification. Qualifying the intended production supplier during the pilot, on lots from the intended plant, therefore belongs inside the work rather than after it. Stock behaves oddly as well: small quantities of many things, short-dated, often held outside the main system, with a real risk that trial material and production material get confused unless segregation is explicit.

Exit criteria, or the pilot that never ends

The characteristic failure is a pilot line quietly becoming a permanent low-output production line, complete with customers, complaints and a batch record system nobody designed. Writing exit criteria at the start — the parameters to be established, the capability to be demonstrated, the consecutive conforming runs required — turns the exercise into a project with an end. Data capture deserves proper systems even at this scale, because trial records form the evidence base for later validation and for regulatory files. Enlarging the pilot is rarely the answer, since that capacity sits on equipment chosen for observation.

Frequently asked questions

How do we stop pilot output being sold as ordinary product?

Decide the release status before the first run and enforce it physically. Trial batches should carry distinct lot numbering, sit in a segregated location and require an explicit release decision naming who may sell them and against which specification. Where the material is genuinely saleable, treat it as a formal product with the same documentation as any other; where it is not, book it to development cost and dispose of it. The ambiguity in between is where findings and complaints originate.

What should a pilot line measure that a production line does not?

Everything the production line will be unable to see later. Temperature and pressure profiles across the whole cycle rather than at set points, sampling at intermediate stages, energy draw, cleaning effectiveness, and how the process responds when inputs are deliberately pushed toward their limits. The purpose is to establish where the operating window ends, not to demonstrate that a nominal recipe works. A pilot that only ever runs at target conditions has answered the easiest available question.

Can a contract manufacturer run our pilot instead of us building one?

Often yes, and it conserves capital, but the arrangement needs care about who keeps the knowledge. Trial data, parameter maps and failure observations are the asset being created, so the agreement should assign that record to you and specify the raw data you receive rather than a summary report. Check as well that the equipment genuinely resembles what you intend to install, because a process proved on somebody else's machinery may not transfer to yours.