The median fault description on our platform is 41 characters long, about the length of "motorlampa lyser" with room to spare. The average is more than double that, 97 characters, and the gap between those two numbers is the whole argument: a small number of long write-ups are dragging the mean up while most technicians type almost nothing.
This comes from 1865 diagnostics that included a written description, out of 2428 total, on Wrenchlane's own platform. It's our own data, not an industry survey, and it only tells you how technicians write into one particular text box. But the pattern is consistent enough that I'd be surprised if it didn't hold elsewhere too.
The median is the honest number
Averages lie when a distribution is skewed, and this one is badly skewed. The 90th percentile description is 226 characters. The longest one on record runs to 3086 characters, a full troubleshooting narrative. Those long entries exist, and they're genuinely useful when they show up, but they're not what a typical job looks like. The average word count across all descriptions is 16.1 words. Most technicians are not writing case notes. They're writing a label.
What "short" actually looks like
Broken into bands, the shape of the data is stark.
| Length | Count | Share |
|---|---|---|
| 1-2 words | 304 | 16% |
| Short phrase (≤ 25 chars) | 283 | 15% |
| One sentence (26-80 chars) | 777 | 42% |
| Detailed (81-300 chars) | 372 | 20% |
| Full case history (300+ chars) | 129 | 7% |
Add the first two rows together and you get 31 percent of every written description that is a single symptom word or a bare phrase, nothing else. "Motorlampa." "Airbag." "Misfire." One sentence with a little context is the single most common shape, at 42 percent, but it's still one sentence. Only 7 percent of technicians write anything resembling a full case history. If a diagnostic tool is designed around the assumption that the technician will describe symptoms, conditions, and prior checks in the way a service manual would, it's built for a shape of input that shows up about a quarter of the time.
The complaint is "the light's on," not a diagnosis
Look at what technicians actually say the problem is, in their own words, and the picture sharpens further. The single biggest category, at 21 percent, is simply that a warning lamp is on: engine light, check-engine, MIL, or any dashboard lamp named as the reason for the visit. Misfire or rough running follows at 10 percent. "A function doesn't work," with no symptom beyond that, is 7 percent. Power loss or limp mode is 6 percent. Won't start is 5 percent.
Notice what's missing from most of these: a hypothesis. The technician isn't telling you what they think is wrong, they're telling you what the driver noticed. That's not a failure of communication. That's the actual state of knowledge at the point the car arrives. Four percent of descriptions name a fault area with no symptom at all, things like "airbag fel" or "felkoder på egr", a system plus the word fault and nothing about what the car actually does. That's a technician typing what they already suspect from a quick look at the codes, not what the customer told them.
Technicians converge on the same handful of phrases
Across 1411 distinct description texts, 241 were written more than once by different people, word for word. "Motorlampa lyser" appears 25 times. "Motorlampa" on its own appears 19 times. "Engine misfire" appears 13 times. "Crank no start" appears 10 times. These aren't templates anyone was handed. They're independent technicians converging on the same short, plain phrasing because that's genuinely how the trade talks about these faults.
The language split reinforces the same point in a different dimension. 54 percent of descriptions are in Swedish, averaging 95 characters, 33 percent in English, averaging 102 characters, with the rest split across Russian, Italian, German, Romanian, Slovak and Danish in smaller numbers. None of these are translations of a manual. They're native, workshop-floor phrasing, in whatever language the person typing happens to work in. A tool that expects a single language, or expects that language to sound like a repair manual, is going to miss most of what actually gets typed.
The blind spots a text box can't fill
Not every entry carries information at all. 3 percent of descriptions are just the scan tool's own wording for a code, copied and pasted, giving no context the code didn't already carry. Another 3 percent are nothing but the fault code itself, no words at all, things like "p0420" or "p0017,p0014,p0089." And 605 descriptions across the whole set didn't match any recognisable complaint pattern, 97 of which were two words or fewer and simply didn't carry enough to categorise in the first place.
That's not a criticism of the technicians who wrote them. If the fault codes already say what's wrong, typing it out again in prose is wasted effort, and everyone in a busy bay knows it. It does mean that any system relying on the free-text description alone, without also reasoning across the actual codes, is working from an incomplete picture on a meaningful slice of jobs. It's one of the reasons we built the diagnostic platform we build to reason across the fault codes and the description together, rather than leaning on either one in isolation.
What to do with this
If you're writing up intake forms, or training a junior service advisor on how to fill one out, stop asking for a paragraph. Ask for the two things that actually show up in this data: what the customer said, in three words or fewer if that's all there is, and what the light or the behaviour is doing. "Motorlampa lyser" is a perfectly good job card entry. It's also 97 percent of what most technicians are going to give you anyway, so a form that demands more than that mostly just gets ignored.
If you're evaluating any diagnostic tool, ask it to handle a two-word input in the technician's own language before you ask it anything else. That's the median case, not the edge case.
I'd be curious whether other shops see the same split between what the intake form asks for and what actually gets written on it.