The Observations a Customer Can Reliably Report
Why this matters
Remote work is only as good as the evidence coming back down the line, and that evidence is not uniformly bad. It is sharply tiered. The same customer who cannot tell you whether a sound is a rattle or a grind can tell you with near perfect accuracy that a light is lit, that the machine started four times in the last hour, and that the floor was dry at seven this morning. Knowing which tasks are reliable lets you spend your questions on the ones that return real data and stop asking for the ones that return noise. There is a companion article cataloging what customers habitually get wrong; this one is the positive side, the ladder of what they get right and how to move a weak report up it.
The ladder
| Tier | Observation class | Why it holds up | Example ask |
|---|---|---|---|
| A | Binary visible state | No judgment, no vocabulary, no scale | "Is the small light on or off, right now" |
| A | Counts | Counting is the one measurement everyone can do | "How many times did it start in the last hour" |
| A | Clock times | People are far better with clock times than durations | "What time did you turn it on, what time did it stop" |
| A | Photographs of static things | The image carries the data, not the description | "Photograph the plate on the side, close enough to read" |
| B | Durations timed with an instrument | Reliable with a stopwatch, unreliable from memory | "Start your phone timer when it kicks on" |
| B | Side by side comparison | Comparison is easy, absolutes are hard | "Does the tap by the equipment fill slower than the far one" |
| B | Order of events | Sequence survives memory better than timing does | "Which came first, the noise or the shutdown" |
| B | Color of a liquid | Coarse categories only, clear versus not | "Is the water clear, cloudy, or colored" |
| C | Magnitude with a stated anchor | Only as good as the anchor you supply | "Can you hold a normal conversation next to it" |
| C | Smell in broad categories | Real signal, poor precision | "Does it remind you of anything, or nothing at all" |
| C | Worse, same, or better over time | Directional, not quantitative | "Compared to a week ago" |
| D | Absolute magnitude | No reference, guesses freely | "How hot is it," "how much water" |
| D | Sound identification | Everyone's words for sounds differ | "Is it grinding or knocking" |
| D | Location inside an assembly | Requires a mental model they do not have | "Is it leaking from the valve or the fitting" |
| D | Naming a component or a state | They will map an unknown word onto the nearest familiar thing | "Is the breaker tripped" |
Tier A and B are evidence. Tier C is a lead worth having. Tier D is conversation, and treating it as evidence is how remote diagnoses go wrong.
The tier is a property of the task, not the person
This is the part people get backwards. An articulate, technically curious customer is not meaningfully better at Tier D than anyone else, because Tier D failures are structural. Nobody has a calibrated internal reference for temperature, volume, or loudness. Everybody's private vocabulary for sounds differs. Nobody can localize a fault inside an assembly they have never seen opened.
What a technical customer does have is more vocabulary, and that actively increases risk, because their Tier D answers arrive dressed in trade words and slip past your filter. "The pressure is low" from someone who owns the word sounds like a measurement and is usually a guess. Ask them for the reading and, if they do not have a gauge, treat it exactly as you would treat "the water seems weak."
The mirror also holds. A customer who is embarrassed about knowing nothing is fully reliable at Tier A and should be told so. "You do not need to know anything about the machine for what I am about to ask. You just need to count."
Moving a report up the ladder
Almost every Tier D question has a Tier A or B version that answers the same underlying thing. This table is the working core of the article.
| What you want to know | Tier D ask that fails | Tier A or B conversion |
|---|---|---|
| Whether it is overheating | "Is it running hot?" | "Hold your hand near it, not on it. Comfortable, yes or no" |
| How much water | "Is it a lot of water?" | "Put a dry towel down for 30 minutes, then tell me if it is damp, wet, or soaked through" |
| Whether flow is reduced | "Is the pressure low?" | "Time filling the same container at two different taps and give me both times" |
| Where a leak comes from | "Is it the supply or the drain?" | Dry it, then three timed windows: nothing used, pressurized only, drain used |
| Whether it is cycling too often | "Does it short cycle?" | "Count the starts in one hour, twice: once in the morning, once in the evening" |
| Whether output has dropped | "Is it weaker than it was?" | "Count run time in an hour now, and again after we change one thing" |
| Whether a part is correct | "Is it the right one?" | "Photograph the markings on the old one and the new one, side by side" |
| Whether it is getting worse | "How bad is it now?" | "Same three observations, same times, one week apart" |
Two patterns run through all of them. Replace judgment with counting or timing. Replace an absolute with a comparison, either against another point in the building or against the same point at another time.
The negative observation is the one you cannot get any other way
The most underrated evidence in remote work is a long-run absence, and only the customer has it. "It has never done it in the first ten minutes." "It has never needed a reset." "It has never happened in winter." "It has never leaked overnight."
A single visit cannot establish any of that. Someone who has lived with the equipment for four years can, and negatives eliminate whole branches at once. Ask for them explicitly, because customers volunteer what happened, never what did not.
Grade them honestly. A confident negative about a memorable event, needing a reset, water on the floor, is strong. A negative about something they would have had no reason to notice, a light being on, is weak, and should be converted into a forward-looking observation instead.
The trust budget
Compliance with observation assignments is finite and it decays. Plan on it.
Roughly, the first assignment in a contact gets done properly, the second gets done, and the third gets done if it is easy. Anything past three in one contact gets guessed at, and a guess that arrives labeled as an observation is worse than no answer, because you will weigh it as evidence.
So the practical rule is a maximum of three assignments per contact, ordered so the highest-value one goes first, each stated in one sentence, and each confirmed back to you before the call ends. Spend the budget on Tier A and B tasks. A Tier D question wastes a slot and returns nothing.
Two things extend the budget. Explaining what an observation will rule out, because people carry out a task they understand the purpose of. And making the first one trivially easy, because an early success builds willingness for the harder one.
Who is at the site changes what is reliable
The ladder shifts with the person holding the phone.
A long-term owner or occupant is your best source of negatives and of change history. They know what is new in the building. They are usually the weakest at doing anything requiring access.
A tenant is often good at Tier A and poor at history, because they were not there for it, and they may be reluctant to report anything that could be read as their fault. Ask about changes neutrally and without any implication of blame.
A caretaker or maintenance person can usually do more physically, which tempts you to push past the safety boundary. Their willingness is not the same as qualification. The boundary is the same for them as for anyone else: floor level, covers on, nothing hot or live.
A trades-adjacent person from another discipline is the trickiest. They can measure and they will use vocabulary confidently, and their vocabulary may not mean what yours means. Take their numbers, ask what instrument produced each one, and check their words the same way you would check anyone's.
The worked example
A customer calls three days after changing a filter themselves: "it is not blowing as hard since I put the new one in, I think the fan is going."
Two Tier D claims in one sentence. The magnitude judgment about airflow, and the component naming. Neither is evidence. Both are useful leads.
The first ask was Tier A: photograph the old filter and the new one side by side, close enough to read the printed markings. That returned an image showing the replacement was a much denser media rating than the one it replaced, which is a real difference and one the customer had no way to evaluate, because the new one was physically the correct size and looked like an upgrade.
The second ask was a count, also Tier A: how many minutes out of the next hour does it run. The answer came back as 45 minutes of run in a 60 minute hour, so 75 percent duty.
The third was the comparison, and here the shop supplied the other half rather than asking the customer to remember it. The service ticket from the same season the previous year recorded a typical pattern near 20 minutes of run per hour, so about 33 percent duty. Seventy-five percent against 33 percent is a jump of more than double, a bit over 2.3 times the run time for the same job.
Three observations, all Tier A, all done in one contact, and none of them requiring the customer to open, touch, or evaluate anything. Together they say the system is working far harder than its own history for the same season, starting within days of a consumable change, with a photograph showing that the consumable is materially more restrictive than what it replaced.
The disposition was to fit a filter matching the previously used rating and repeat the same count. That came back at about 20 minutes in the hour, matching the historical record.
Note what was never asked. Nobody asked whether the airflow felt weaker, because that is Tier D and it was already the customer's own claim. Nobody asked whether the fan sounded strained, because sound identification is Tier D. The entire case ran on a photograph, two counts, and a record the shop already held, and it closed without a truck.
How to verify a customer report before you act on it
Ask the same fact twice, framed differently, and compare. A number that changes when you rephrase was never solid.
Check it against something you hold. Your own service history, a previous ticket, a photograph from a past visit. Internal records are the cheapest cross-check available and they are consistently underused.
Test internal consistency. If they report a symptom that runs continuously and also report the machine only runs occasionally, one of the two is wrong. Chase the contradiction rather than picking the answer you like.
Weight by tier, in writing. Note the tier next to each observation in the record. When a case later collapses, it is nearly always because a Tier C or D answer was carrying weight it could not hold, and the annotation makes that visible immediately instead of after the second wasted visit.
References
- Trade-standard practice for structured fault reporting and evidence documentation
- Manufacturer documentation on consumable specifications and the effect of substitutions on system load
- See related: What a Customer Consistently Gets Wrong Describing a Fault Remotely
- See related: How to Ask a Non-Technical Customer a Diagnostic Question
- See related: What You Can and Cannot Conclude Without Being On Site