Starting a 3D print farm: the first ten decisions

A print farm is not a hobby with more printers. It is a small manufacturing business, and almost every problem people hit in the first year traces back to a decision made in the first month, usually without noticing it was a decision.

Here are the ten that are expensive to change afterwards.

1. What you are selling

Three businesses hide under "3D printing services", and they want different machines, different filament, and different working hours.

Your own product, printed in quantity. Long unattended runs of the same file, one or two materials, and the hard parts are demand and finishing rather than printing.

Made to order for other people's files. Every job is new, quoting is constant, and your margin lives or dies on how fast you can price a job accurately.

Prototyping and engineering work. Fewer jobs, higher value, tighter tolerances, and customers who ask about materials you have not bought yet.

Most farms drift into a mix. That is fine, but pick the one you are optimising for, because the answers below change with it.

2. One printer model, or several

One model is cheaper to run than it looks. One spare parts drawer, one set of profiles, one set of quirks to learn, and any job can go on any machine - which is what actually lets you fill the queue.

Several models buy you range: a bigger bed, a hotter hotend, a material one machine will not touch. They also multiply everything else, and the second model is where a farm quietly stops being interchangeable.

Start with one, and add a second model only when a job you are turning away pays for it.

3. How many, and when

The instinct is to buy printers until the queue clears. The better rule is to add a printer when the ones you have are running at around seventy per cent of the hours you are willing to run them - and not before.

Below that, a new machine adds cost and no output. Above it, you have no room for a failed print or a rush job, and every delay reaches the customer.

Work out what a printer actually costs you per printing hour before you buy the next one. The printer hourly rate calculator takes the purchase price, the hours you expect to get out of it, the power it draws and what you spend keeping it alive, and gives you the number that belongs in every quote.

4. Where they go

Printers are noisy, warm, and they smell. They also do not like draughts.

The practical constraints, in the order people discover them: the noise carries further than you expect and it runs at night; enclosed machines need the room ventilated, particularly for anything that is not PLA; a heated bed and a hotend on every shelf adds up to a serious load on one circuit; and a machine sitting under a window will produce warped prints all winter and you will blame the filament.

Shelving beats a bench once you pass about four machines, and rigid shelving beats cheap shelving, because a wobbling shelf is a layer shift.

5. Power, and what happens when it goes

Add up the running draw of every machine and compare it to the circuit, not to the socket. Measure one printer with a plug meter over a full print rather than trusting the nameplate, because the nameplate is the peak.

Then decide what happens when the power blips. Some printers resume, some do not, and the difference between them is a whole night of production. If your machines do not resume, an unattended overnight queue is a bet you are making every night.

6. Filament: how much, and how dry

Buying in bulk is cheaper per kilogram and worse for cash flow, and the crossover is not where people put it. Twenty kilos of a colour you stop selling is not a saving.

Dryness matters more than most new farms expect, and it matters more the more humid your room is. PLA tolerates neglect; PETG, nylon and TPU do not, and the symptom is not obvious - stringing, weak layers, a rough surface that you spend an evening blaming on the nozzle.

Whatever you do, decide now that a spool has a cost and not just a colour. A farm that cannot say what a print cost in filament cannot say what it cost.

7. How a job gets onto a machine

This is the decision nobody makes and everybody ends up with.

The default is a person, a memory, and a phone. It works at three printers. At eight it starts dropping jobs, and the symptom is a customer asking where their order is and nobody being certain.

You need one place that says what is queued, what is printing, what is done, and which machine each job is on. It does not have to be software. It does have to be one place, and it has to be the place everyone looks - a queue that exists in two places is a queue that exists nowhere.

8. What a print costs you, before you quote it

Quote from cost, not from feel. Filament used, machine hours, labour, a waste allowance, and a multiplier.

The trap is that every one of those numbers drifts. Filament prices change, the machine that cost thirty cents an hour last year has printed twice as much as you assumed, and the labour figure was always a guess. A price list written once and never revisited is a price list slowly turning into a discount.

9. Who does the finishing, and when

Support removal, sanding, gluing, painting and packing do not get faster when you add a printer. They are the part of the business that scales by hiring, and the part that quietly eats your margin if it is not on the quote.

Time it on a few real jobs. Almost everybody underestimates it, because what they remember is the job that came off the plate clean.

10. How you know whether the month made money

Revenue is not profit, and a busy month is not a good month. The farms that survive their second year are the ones that can answer three questions:

• What did each order actually cost to produce?
• Which customers and which products make money, and which ones only make work?
• What is sitting in filament on the shelf, and what is it worth?

You can answer these with a spreadsheet, and plenty of shops do. The spreadsheet stops working at about the same point the phone-and-memory queue does: it does not know a spool ran out, it does not decrement anything when a job prints, and it only tells you the truth on the evening somebody sits down and updates it.

That is the point of what we build - orders, printers, filament stock and costing in one place, where finishing a job updates the stock and the margin without anybody retyping it.

The one to get right first

If you only act on one of these, make it number seven. A farm with an unreliable queue loses customers it never finds out about, and every other problem on this list is visible in a way that one is not.