Six companies in the world can build an inkjet printhead from scratch: Epson, Ricoh, Konica Minolta, Xaar, Kyocera, and Fujifilm Dimatix. That single fact settles most of the confusion around cheap parts, because if only six companies make heads, then a head sold as compatible cannot be a second-source product. It is either a head pulled from a scrapped machine, cleaned, and resold as new, or a head built for one ink chemistry wearing the housing of another. Both fit. Both run. One of them quietly ends your week. The practical rule for a working shop is to stop asking whether a part is OEM or compatible and start asking which of those three things you are actually being sold. What follows are the six checks to run before any non-original part goes into a production machine. Do them in order. The first three cost nothing but ten minutes at the machine, and they catch most of the expensive mistakes.
The Word Compatible Covers Three Different Products
A damper ordered from an aftermarket factory is a real product, made by a real factory that has been molding them for years. A printhead sold as compatible is not that. Nobody outside those six companies has the tooling to fabricate a piezo head, so the part in the box is one of two things: a head pulled from a scrapped machine and resold, or a head built for one ink chemistry being sold as the version for another. The first may work for a month or a year with no way to tell from the outside. The second works until the wrong ink reaches the nozzles, and then it does not work at all.
That distinction changes what you should be shopping for. For dampers, cap tops, wipers, cleaning cartridges, and tubing, the aftermarket is a legitimate second source, and the question is which supplier, not whether. For printheads, ink pumps, carriage boards, and main boards, there is no second source. There is the original part, and everything else is a story about where it came from.
Run one test on any listing before you add it to a cart. Ask what the part physically is, not what the label calls it. If the answer is that a factory builds it to a specification, you are buying an aftermarket part and the specification is what you need to see. If the answer stays vague about origin, you are buying a pulled part with an unknown service life, and the price reflects that risk whether or not the seller admits it.
The cost of getting this wrong is not the price of the part. It is the hours the machine sits idle while you wait for a second part, plus the media you scrap when the first one fails mid-roll, plus whatever you already promised a customer for the week. A shop running one production machine has no slack for that arithmetic.
Check One: Read the Part Number Off the Failed Part, Not the Invoice
The invoice tells you what you bought last time. It does not tell you what is in the machine now. Those two things drift apart, because machines get serviced by different people, parts get swapped between identical units, and a previous owner may have fitted something that fits but was never the specified part. When a replacement goes in and misbehaves, the number that matters is the one molded or printed on the part you just removed.
Photograph it before it goes in the bin. The number is often small, often on a face that ends up hidden once installed, and often accompanied by a second string that identifies the production batch rather than the part. Both are worth having, because the batch string is what a supplier needs when you report that a run of parts failed early.
Model generation names are not part numbers. Printer families carry a name on the plate and then ship with different heads, different dampers, and different capping assemblies across the production run. A head that fits a machine with the same model name can be the wrong head for your serial number range. This is the most common way a part arrives that looks correct and prints wrong.
Do not try to settle this with a caliper. Two parts can measure the same on every dimension you can reach and still differ in the internal filter, the spring rate of the valve, or the hardness of the seal. The number is the only reliable identifier here, and reading it takes less time than measuring.
Do this today: make a card for the machine and tape it inside the front cover. Four lines. Full model string with any suffix, firmware revision, the part numbers of the head, the dampers, and the cap top currently installed, and the date each was fitted. Ten minutes of work, and every future order stops being a guess.
Check Two: Match the Ink Chemistry Before the Shape
Two printheads can be dimensionally identical and still be different parts. The housing, the connector, and the mounting holes do not change between a head rated for water-based ink and one rated for solvent. What changes is inside, in the materials that touch the ink and in the waveform the head expects. Fit the wrong one and it may print for a day before swelling, corrosion, or thermal drift starts to show.
Suppliers who know their stock will state the ink type in writing. Suppliers who are moving boxes will state the printer model only. That difference is the whole test. If a listing says it fits a set of machine models but never says which chemistry, treat it as unknown until somebody confirms it in a message you can save and forward. Ask for the ink chemistry in the same message where you ask for the price, and keep the reply filed with the invoice.
Cleaning solution belongs to the same question. A solution matched to the wrong chemistry will not dissolve the ink it is supposed to lift, and flushing a line with it repeatedly leaves the line worse than it started.
White ink needs its own paragraph. Pigmented white carries a heavy solids load that settles out of suspension, and the damper on that channel takes the worst of it. Aftermarket white dampers are usable, but they load up faster than the same damper on a color channel, and a damper that has stopped passing white cleanly starves the head while the nozzle check still looks plausible. Change white dampers on a shorter interval than the color set, and after fitting a new one run more than one cleaning cycle before you judge it. On most machines with a white circuit, three cycles is the number that brings the channel back to a stable nozzle check.
Shop floor action: keep white and color dampers in separate bags with the fitment date written on each bag in marker. Same part number, different service life, and the bag is what tells you which is which at eleven at night.
Check Three: Confirm the Board and Firmware Revision
This is the check that catches the failures nobody blames on the part. The same machine model ships with different controller boards across a production run, and the board decides what the firmware expects to see. A part tested against one firmware revision can be rejected, or half accepted, by another, and the machine reports it as a fault rather than as a part mismatch.
Read the firmware revision from the machine, off the control panel or the service menu, not from the manual and not from the supplier page. Then read the revision marking on the part you are removing. Then ask the supplier which firmware revisions their part has actually been tested against, in writing. Three readings, ten minutes, and the answer is either a document or a shrug.
Watch for the quiet version of this failure. The machine accepts the part, prints, and then stops early or throws an identity fault partway through a run. Operators report that as an intermittent machine fault and it gets escalated as one. The faster test is a swap: fit a part from a batch you already trust. If the fault follows the part, it is a batch problem and it belongs in a conversation with the supplier, not in a service call.
When you do contact the supplier, send four things in one message: the full model string, the firmware revision, the reference of the failing part, and the batch code. A supplier who asks for exactly those four is used to solving this. One who answers with a price and a shipping date is not.
Keep that firmware reading on the machine card too, and update it whenever the board is serviced. Changing firmware on a machine that is running fine is a separate argument from this one. If nothing is broken, the safer position is to record the revision you are on and ask any supplier to match it, rather than to change the machine to match a part you already bought.
Check Four: Price the Downtime, Not the Part
Two parts, one at half the price of the other. The cheap one is not cheaper if it costs you a day. Work out your own number before the next failure, because you cannot do this arithmetic while the machine is down and a customer is on the phone.
Start with what an hour of the machine is worth to you: the billing rate on the work it would otherwise be running, or the wholesale value of the output if you sell through. Multiply by realistic hours lost, which for a part that fails at the wrong moment includes waiting for the replacement, not just fitting it. Add the media scrapped in the job that was running. Add expedited freight if the schedule forces it. The total is the real price of the cheap part, and it is usually several times the difference you were trying to save.
Now apply the same math to parts where the downside is not downtime but damage. A capping station that does not seal properly never fails loudly. It lets the head sit against air, the nozzles dry out, and the head that was fine last week stops firing on a block of channels. The saving on the cap is trivial next to the head. If you cannot verify the rubber compound and the seal tolerance on an aftermarket capping station, buy the one you can verify. Note that this reasoning applies to the cap and not to the wiper, which is a simpler part and far more forgiving of a good aftermarket copy.
Write your hourly number on a card and keep it next to the purchasing login. Then the rule becomes mechanical: if a part can plausibly cost more than one shift of production, the price difference stops being the deciding factor.
Checks Five and Six: Set a Test Window and Keep One Known-Good Part
Any non-original part that goes into a production machine gets a test window, decided before installation rather than after something goes wrong. 72 hours or the first full roll, whichever comes first, is a workable default for most shops. During that window the machine runs its normal work, and every nozzle check gets compared against the baseline you printed before the change, not against a memory of what it usually looks like.
Keep one known-good part on the shelf. Not a spare to fit, an instrument to test with. When the machine misbehaves after a change, swapping in a part you trust tells you in twenty minutes whether the new part is the problem, and that single answer decides whether you are on the phone with a supplier or with a service tech. It is the cheapest diagnostic in the building, and most shops do not have it because the good part got used as a spare.
Log three lines when a part goes in: date, roll count at fitting, and the result of the first nozzle check after installation. Three lines per part, kept on the machine card, is enough to show that a batch of dampers is averaging two months instead of six, which is a supplier conversation you want to have with evidence rather than irritation.
During the test window, do not schedule the job you cannot reprint. Run routine work, watch the nozzle check, and keep the deadline work on the other machine or hold it until the window closes. A part that is going to fail usually does it inside the first roll, and you want it to fail on a job you can afford to lose.
One commercial reality to plan around: printheads are sold final sale across the industry, and for good reason, because once a head has been connected nobody can separate a manufacturing defect from a short caused during installation. That policy makes the pre-purchase checks your only protection. Confirm fitment, confirm chemistry, confirm the firmware revision the supplier tested against, and get all of it in writing before the box ships.