Start with the run you are actually asking for
A plan that starts from the machine list will not survive the first week. It should start from a description of the run: which item, in which form, arriving from where, leaving as what, and how many times the line has to repeat that sequence before the order is satisfied. That sentence is the anchor for every later decision about fixtures, queue positions, hold points and staffing. If the description cannot be written in one paragraph, the order is not yet defined, and a plan written on top of it will be rewritten anyway.
For a line built around gallium arsenide production line equipment, the run description also fixes the substrate format. The handling rules, the usable area of a fixture and the position of inspection steps all follow from the format that is arriving. Two orders that look similar on a purchase requisition can describe different formats, and a plan copied between them will fail quietly rather than loudly.
Write the material record before you write the schedule
Small runs tend to carry more incoming variation than steady production: different supplier lots, mixed carrier types, parts that arrive with documentation gaps. The plan should therefore open with a material record that states, for each input, what identifies it, what proves it is the right input, and what happens if that proof is missing. A record that says "as received" is not a record; it is a deferral.
This is also where long lead items are separated from short lead items. Anything that cannot be replaced within the planned window deserves its own line, with the trigger for ordering it stated in the plan rather than remembered by whoever has always ordered it.
Leave the wet-process parameters as fields, not as prose
Where the run includes a chemical step, the plan should carry blank fields: solution identity, strength, temperature, contact time, agitation, rinse sequence, and who recorded each value. Do not fill those fields with numbers copied from a datasheet, a forum post or a neighbouring process. A plan that carries unfilled fields is a plan that will be completed by the people who own the process; a plan that carries guessed fields is a plan that will be argued about.
Blank fields also protect the next run. When a process is repeated, the question is usually not what the settings were in principle, but what they were on the day, on that bath, with that lot of material. A filled-in field with no name against it answers neither question.
Handle and layout: the part most plans forget to plan
Write down how the part is held at each station, who changes that holding, and where parts sit between stations. Fragile processed surfaces are damaged far more often in transit between two machines than inside either machine. A plan that names a carrier at every station, names the parking place between stations, and names the direction a part faces when it is parked will save more than any extra inspection step added later.
Then draw the flow once, on one page, showing queue positions. If a queue position has no owner and no maximum dwell time, mark it as an open item in the plan instead of drawing it as if it were settled.
Safety, access and permissions are planning items
Chemical handling, extraction, protective equipment, interlocks and room access decide what a small run can do on a given day. Put those items in the plan with the name of the person who grants access and the condition under which access is granted. A run that stops because nobody was authorised to open a cabinet is a planning failure, not an operator failure.
Include one line on what happens when an enclosure must be opened outside its normal purpose — for maintenance, for a trial, for a sample. That line prevents the improvised workaround that later becomes an incident report.
Plan the changeover, and consider two runs
A small run usually shares its line with something else. The plan should state what has to be removed, cleaned, verified and replaced before the line is allowed to carry this item, and how the line returns afterwards. Where the changeover is long, compare that time against the run length: a changeover longer than the run itself is a scheduling decision, not a technical detail.
Where practical, plan two passes: one whose purpose is to prove the flow, and one whose purpose is to produce. The first pass answers layout, holding and queue questions with cheap consequences. The second pass carries the commercial intent. Trying to do both in a single pass is the most common way a small order misses its date.
Change control: who may alter what, and where the alteration is recorded
Small runs change constantly — a substitute carrier, a shortened dwell, a reordered inspection. The plan should not pretend otherwise; it should say who may authorise a change, how the change is written down, and which steps may never be altered without a formal decision. A blank change log printed alongside the plan gets filled in. A policy that exists only in an email thread does not.
Where automation carries part of the run, as in many automated production line solutions, also state who may touch a program and who must sign off a restored version. Program edits made in the middle of a short run are frequently invisible afterwards unless they are recorded.
Evidence: what the run must be able to prove afterwards
Decide, before the run, which records it must leave behind: identity of the material, settings used, who performed each step, inspection results, and any deviation with its justification. Working backwards from the report the customer will ask for is faster than collecting everything and hoping. Evidence that is reconstructed after the fact carries far less weight than evidence captured in the moment.
Questions to put to the equipment supplier in writing
Before the plan is frozen, send a short list: which formats the equipment holds, what changeover it requires between those formats, what services it expects from the room, what it does when a step is interrupted, which faults are logged and how they are read, what consumables it needs and at what rate, and what training or support comes with it. Ask the same questions of every vendor so the answers can be compared.
For smaller or more specialised cells, including thin-film solar equipment, add one practical question: how the vendor demonstrates the process on a customer's own material, and what is agreed before that demonstration happens. A demonstration arranged on the vendor's standard sample tells you far less than a demonstration arranged on the format you will actually run.
Two planning mistakes worth naming
The first is treating the plan as a document for the customer rather than an instruction for the line. A plan nobody on the line can read in a few minutes will be ignored, and the run will be improvised. The second is planning only the line and not the waiting: approvals, incoming checks and external services routinely consume more of a short run's calendar than the processing does.
A one-list way to close the plan
Read the plan one last time and ask, for every line: is there a name against it, or is it blank on purpose? Names make a plan executable. Blanks make it honest. A line that is neither — vague, unowned, and filled with plausible values — is where a short run loses its schedule.
