The Documentation a Manufacturer Never Publishes

Why this matters

Techs treat missing information as an accident, as though the right search would eventually turn it up. Most of it is missing on purpose, for reasons that have nothing to do with you and will not change. Once you know which categories are structurally absent, you stop burning an hour a week hunting for documents that do not exist, and you start building the substitute. The shop that builds substitutes answers questions its competitors cannot answer at all: which part fails first on this family, how far out of spec is still fine, and how long this machine really lasts around here.

The four reasons something never gets published

Every gap traces back to one of these, and knowing which one you are up against tells you whether a substitute is even possible.

• Liability. Anything that reads as an admission gets written out. Failure rates, known defects, and the real service life of a component are all statements that could be quoted back in a claim. This is why a service bulletin describes a corrective action and rarely the population that needed it.
• Competitive value. Failure and warranty data is expensive to collect and useful to a competitor. It is an asset, not a document.
• It does not exist in a publishable form. A manufacturer knows what its equipment does on a test stand under standardized conditions. It does not know what your climate, your water chemistry, your dust load or your local installation practice do to it. Nobody knows that except the service organizations working in it.
• The audience is not you. Some of it is written and distributed, just through a dealer or authorized-service channel that a general shop cannot reach.

That last one is the only category where more searching helps.

Where the substitute actually lives

The pattern across the whole table: the substitute is almost always something you generate, not something you find. Which means the shops that have it are the ones who decided three years ago to start recording.

Worked: building the failure-frequency document nobody will sell you

A shop pulled its own service history on one equipment family over three years: 84 repair visits.

The raw distribution came out as 31 visits attributed to one component type, 22 to a second, 14 to a third, and 17 spread across everything else. Those add to 84. As percentages of the raw total, that is about 37, 26, 17 and 20 percent.

Then they did the step that separates a useful document from a misleading one. Of the 84 visits, 9 were on units the shop had installed itself and the fault was an installation defect, not a component failure. Those do not belong in a failure-frequency document at all, because they measure the shop's install process rather than the equipment. Removing them leaves 75 equipment-failure visits.

Five of the 9 had been sitting in the leading component's bucket. So the leading component corrects from 31 to 26, which against the corrected base of 75 is about 35 percent. Two more of the 9 had been in the second bucket, correcting it from 22 to 20, which is about 27 percent of 75. The remaining 29 visits, which is 75 minus 26 minus 20, cover everything else at about 39 percent.

Note what happens to the comparison if you are careless here. Reporting the corrected 35 percent against the original 37 percent raw figure is not a like-for-like comparison, because the 37 percent still has installation defects inside it. Once you correct the numerator you correct the base too, or you say plainly that the number you are comparing against is uncorrected. That is the single easiest way to make a small correction look like a trend.

What the shop did with it. The leading component at about 35 percent of equipment-failure visits went onto the truck as stock, along with the second at about 27 percent, which together account for a bit under two thirds of all failure visits on that family. The third bucket did not, because a component that shows up in well under a fifth of visits ties up truck space that a faster mover needs.

And the second finding, which was the more valuable one. Those 9 installation defects were about 11 percent of all visits on the family, and they were the shop's own work. That is a training and checklist problem the manufacturer's documentation could never have surfaced, and it only became visible because someone insisted on separating install defects from equipment failures instead of counting all 84 as failures.

The honest error bar. Seventy-five visits in one climate over three years is a small sample. Treat the ORDER as reliable, because a component appearing in a third of visits and one appearing in a sixth are genuinely different. Treat the PERCENTAGE as directional: the same component in a much larger population could sit meaningfully above or below 35 percent. Do not quote a figure like that to a customer as though it were a manufacturer statistic.

The one that is published but not to you

Service bulletins are real documents, written by the manufacturer, describing a specific problem and its corrective action, and distributed through a dealer or authorized-service channel. If you can get access to that channel for the families you work on most, do it, because a bulletin is the only place where the manufacturer says out loud that something was wrong.

Where you cannot, the counter is the substitute. Ask two questions on every part order for an older machine: has this part number superseded, and does the new part look or measure different from the old one. A part that superseded once is normal lifecycle. A part that superseded and physically changed, or grew a bracket, or gained a different rating, is the manufacturer fixing something without saying so. That is inference rather than evidence, but it is inference from a hard fact, and it beats a forum post.

Where a substitute is not a substitute

Substituting works for population-level knowledge and for trending. It fails, sometimes dangerously, for design values.

• Never take a spec from a near model. A neighbouring model number in the same family may differ in exactly the parameter you are looking up. On anything in a fuel train, an orifice size or manifold pressure taken from a sibling model can produce incomplete combustion and carbon monoxide in an occupied space, so where the appliance-specific value cannot be obtained, the appliance does not run.
• Never take a protective-device setting from anywhere but the equipment or the listing. Relief settings, limit setpoints and overload sizes are the last line between the machine and the building. An adjustable device with an unknown setting gets replaced with one of a known rating rather than adjusted until the machine runs.
• Never treat another shop's numbers as your data. Different climate, different water, different install crew. Their order of failures is worth reading. Their percentages are not yours.
• Never let a substitute outrank a measurement. A baseline from two years ago is context. What your meter says today is evidence, and when they disagree the meter wins.

How to keep your substitute documentation from rotting

A shop-built document decays faster than a manufacturer's, because nobody owns it by default. Four habits keep it alive, and they cost minutes rather than hours.

• Record the cause category at close, not at the end of the year. Reconstructing 84 visits from memory produces a document that measures what the office remembers. Capturing the category as each ticket closes produces one that measures what happened.
• Separate install defects from equipment failures at entry. This is the correction that mattered most in the worked example, and it is trivial to do at the moment of the visit and painful to do afterwards.
• Date every derived number and name its source. "About 35 percent of equipment-failure visits, 75 visits, three years to this spring." A number without a base and a date cannot be checked, cannot be updated, and will be quoted years after it stopped being true.
• Re-run it on a fixed cadence. Annually is enough for failure frequency. The value is in the change: a component that moves up the order between two runs is telling you something about a production change, a new install cohort, or your own workmanship, and only a repeated measurement can show it.

References

• Manufacturer service bulletins and part supersession notices, where dealer or authorized-service access is available
• Certification listing label for the standard of construction and required protective devices
• Commissioning and air or water balance reports for installation-specific acceptance values
• Your own service history, parts usage and callback records as the primary substitute source
• See related: How to Work a Job With No Manual at All; How to Document Tribal Knowledge So It Outlives the Person