Early-Life Failure and DPPM
Claims That Cannot Be Ignored

The Bathtub Curve, Priced

Every printer line fails on the same curve, and only one part of that curve is actionable for a distributor. The three regions are familiar: a short early-life band where failures come from marginal components and assembly errors, a long middle where failures are essentially random and driven by use, and a wear-out tail where the parts that were always going to fail eventually do. The middle and the tail belong to the customer's operating environment and to time. The early band belongs to the distributor, because a failure in the first weeks is a failure of either the screening the distributor did or the batch the factory shipped.

Treating that early band as someone else's problem is expensive. A printer that fails in the first month generates a warranty claim, a return shipment, a replacement unit, an unhappy customer and, most damaging of all, a reputation for unreliability in the reference the distributor was counting on. The same failure caught by a burn-in routine before the unit leaves the warehouse costs a day of screening time. The arithmetic is not close, which is why a defined screening step should be standard on every unit a distributor ships, not reserved for the ones that look doubtful.

Burn-In: What To Run Before the Printer Ships

Burn-in in the additive context is not an endurance test and should not be treated as one. Its job is to force the marginal units to fail in the distributor's warehouse rather than at the customer site, and that means exercising the subsystems that carry the highest early-life risk within a window short enough to be commercially practical. Thermal behaviour, motion integrity and extrusion calibration cover most of it.

The thermal soak is the step most often cut and the one that catches the most infant failures. A marginal heater cartridge, a thermistor with a bad crimp, or a board connector that has not seen full current all reveal themselves in the first hour at temperature, and none of them will show up on a unit that is merely powered on and idle. Running the hotend at working temperature for the better part of an hour, with a live thermistor reading watched for drift or dropouts, screens out a class of failure that would otherwise appear in the customer's first week. The reliability metrics that the customer will eventually measure against are set out in our guide to MTBF, MTTR and uptime economics, and the values that return from these screens feed straight into them.

DPPM Arithmetic You Can Do On A Napkin

Defective parts per million, DPPM, is the number that makes field quality comparable across batches, lines and suppliers, because it normalises a failure count to the number of units exposed. It is calculated as the number of defective units divided by the total units shipped, multiplied by one million, and it counts units rather than failure events, so two failed fans in the same printer are one defective unit for the purposes of the metric.

The arithmetic is trivial, and it is the reason a distributor can argue a claim with a factory without a statistics package. Suppose a batch of 2,000 units produces 8 field failures in the first 90 days. The DPPM is 8 divided by 2,000, times one million, which is 4,000 DPPM. That is a reportable early-life result, and it says something useful only after it is broken down, because the same 8,000 DPPM on a smaller batch can disappear inside a healthy aggregate.

The aggregate line is the trap. Averaged across 10,000 units the fleet looks like a respectable 2,600 DPPM, and that average conceals a batch running at 8,000. This is why DPPM has to be reported per batch and per failure mode rather than as a single fleet number, and why the serial list behind the arithmetic matters as much as the count. A distributor that reports only the aggregate will find the factory answering a claim with the fleet figure, and the argument will be lost on presentation rather than on facts.

Separating Infant Mortality From Wear-Out

The classification that decides whether a failure is the factory's responsibility or the customer's is a matter of three questions, and they can be answered from the warranty records a distributor already keeps. Getting the classification right is what prevents two opposite errors: claiming factory responsibility for a wear-out failure and damaging the relationship, or absorbing a batch problem that was the factory's to fix.

• When did it fail, in operating hours? Failure clustered in the first tens to low hundreds of operating hours points to infant mortality. Failure after thousands of hours, spread evenly, points to wear-out or to the customer's process.
• Is it one batch or spread across batches? A failure mode concentrated in units from a single build batch points at that batch and is the factory's to answer. A failure mode appearing at a low rate across all batches points at the design or at a component common to all of them, which is a different conversation.
• Is it one component or several? A single component failing early in a cluster is usually a supplier lot problem. Failures spread across unrelated subsystems point at assembly or shipping rather than at any one part.

The answers map to action: clustered in time and batch means pursue a claim; spread in time and batch means improve the screening or the component specification; failures concentrated in one subsystem across all batches means escalate to the factory as a design issue and expect a revision. Recording operating hours and serial numbers at the point of claim is what makes this classification possible, and it is the same record-keeping discipline described in our guide to fleet asset registry and life cycle records. A cluster of early hotend failures is a concrete illustration of the component-level view, covered in hotend failure modes and escalation.

The Escalation Ladder to the Factory

Escalating a quality problem to an overseas factory works or fails on the evidence pack, and the pack has to name the population, not just the incident. A single failure complaint can be answered with "no failure found" and the conversation ends. A claim that identifies which units are affected, when they were built, and what the failure rate is against the exposed population gives the factory a problem it has to own. Five rungs take it from a support ticket to a batch claim.

• Rung one, support ticket. A single reported failure with the serial. The factory's response is diagnostic advice, and this rung does not establish a claim.
• Rung two, failure report with evidence. Photographs, the failure mode, the operating hours and the affected subsystem. This is where most distributors stop, and it is why most quality conversations stall.
• Rung three, population claim. The serial list, the build dates, the batch identifier, and the DPPM for the affected population. This rung turns an anecdote into a rate, and it is the first one the factory's quality function has to formally respond to.
• Rung four, containment request. A request to hold shipment of the affected batch or component lot pending investigation, together with the distributor's own containment on stock in hand.
• Rung five, root cause and corrective action. A requirement for the factory's 8D or equivalent, with a design, process or supplier change identified and a verification plan. This is the rung that prevents recurrence.

The mistake that keeps distributors stuck at rung two is presenting failures one at a time, as they arrive, so the factory never sees the rate. The rate is the argument. A factory can reasonably decline a single complaint; it cannot reasonably decline a documented claim that 8,000 DPPM of a specific batch share a common failure mode, particularly when the distributor's own incoming records show the batch and the serial range.

What Makes a Claim Unignorable

Four pieces of evidence move a claim up the ladder, and their absence is what lets a factory respond with a request for more information and let the matter lapse. The first is a serial list with build dates, because it defines the population. The second is a failure mode that is consistent across those units, because a consistent mode cannot be dismissed as customer misuse. The third is a quantified failure rate against that population, which is the DPPM figure. The fourth is the distributor's own test data showing the failure reproduced on the bench, because reproduction removes the "cannot verify" answer entirely.

The omissions that most often weaken a claim are the boring ones. Field reports without operating hours, photographs without a scale or a serial in frame, failed returns scrapped before the failure could be examined, and replacement units shipped before the failed one was inspected. Each of these removes a piece of evidence that would have settled the argument, and each is preventable with a simple rule at the point the customer reports the problem.

Feeding Field Data Back Into Product Selection

The end use of a DPPM programme is not the individual claim, it is the next product decision. A distributor that tracks failure rates by component and by batch across a year accumulates something no competitor can buy: a view of which parts of which printers actually fail in the field, at what rate, and after how many hours. That view should drive two decisions. It should determine which components the distributor specifies for the private-label or OEM units it commissions, and it should determine which units the distributor is willing to stock in volume.

Both of those decisions run through the sourcing conversation, and a distributor arriving at that conversation with field failure data arrives with far more leverage than one arriving with a price target. The negotiation mechanics are covered in our guide to OEM pricing negotiation with Chinese factories, and the reverse-logistics side of handling the returns that the failures generate is covered in warranty returns and reverse logistics. A distributor running a burn-in screen and a per-batch DPPM report has, in effect, built a quality function that turns the installed base into a specification document, and that document is worth more than any price list the factory will send.

Bottom Line

Early-life failures are the only part of the reliability curve a distributor owns, and the tools for managing them are unglamorous: a two-to-four hour burn-in screen with a proper thermal soak, DPPM reported per batch rather than in aggregate, a three-question classification that separates infant mortality from wear-out, and a five-rung escalation ladder that arrives at the factory with a population and a rate rather than an anecdote. The commissioning test that establishes the unit is sound in the first place is covered in our guide to commissioning and acceptance testing. Distributors who run all four of these turn field quality from a cost centre into a purchasing advantage, and they do it with records they are already keeping.

← Back to Blog