Name one decision first
Write what the trial is intended to decide: compare formulations, compare proposed thermal conditions, assess repeatability, or produce samples for analysis. State the measured response and the comparison basis. Equipment configuration should support that decision rather than dictate an experiment with no clear endpoint.
Separate the run into observable states
Record preparation, startup, transition, stable collection and shutdown as different states. Define what evidence allows a collection window to begin and what invalidates it. The selected recipe is an instruction; the measured run record is evidence of what actually occurred.
Allocate material before preparing it
Use explicit quantities for startup discard, all transitions, total steady output, terminal circuit material and other losses. Then show which part of steady output becomes the labelled sample set. Unknown machine losses need confirmation; leaving them blank is more honest than silently treating them as zero.
Close the loop with a run packet
Keep formulation revision, equipment configuration, channel definitions, measured data, deviations and sample IDs together. A second operator should be able to understand which samples can be compared and where a condition differed. Use that packet to frame the next configuration or trial question.