An unanswered customer question is one that gets answered at an address the business cannot see, and four fault patterns name those addresses. The record model decides all four: a question that has nowhere to land in the record produces no answer inside the business. Nobody takes responsibility for an answer produced outside, so nothing can be corrected; invisibility is part of the fault’s definition, not a consequence of it.
An unanswered question does not disappear
Two very different faults produce the same quiet. In one of them the request never forms. Nobody calls, no form is submitted, no record opens, and the business is right to say there is no demand. Sorting one kind of request from another belongs to that side of the line, and the article on separating appointments from service requests works through it in full.
The pattern here behaves differently. The question does form. The customer asks it and gets something back. The only trouble is that the place where the answer was produced is not the place where the business keeps records. From outside, that looks identical to silence: no new row, no new message, no new call. Underneath, a question was asked, answered and closed. What is invisible is not the question. It is the address.
Tahlil.com carries the claim this article rests on. A lab report handed to a patient on paper is a step in a process, and that step was never converted into software. The paper is handed over, the patient walks out of the door, the step ends there. Saying where a step ends is a diagnosis. Saying what replaced it is a different piece of writing, and not this one.
The same fault repeats across builds that have nothing to do with each other: a health platform, an industrial classifieds platform, a perfume store. In all three the customer’s question finds an answer without ever touching the system, and what separates the patterns is the address that question travels to rather than the sector. More than one of them can run in the same business at once, and closing one leaves the others open.
A report that takes medical training to read creates a question
Tahlil.com is a health platform that connects a patient’s lab values to a specialist doctor’s interpretation. The fault this article follows starts earlier than the platform, at the laboratory door.
The report a patient receives consists of reference ranges and abbreviations. Nothing on the page is missing. The value is printed. The range is printed. The unit is printed. Reading the page still takes medical training.
Written and readable are not the same property, and the gap between them is where this fault lives. The report is not an answer; it produces a question in the patient’s hand. The work does not finish when the paper is handed over, it starts there.
This is not the familiar problem of an incomplete document. The document is complete, and detailed beyond what most readers need. The problem is that it was never composed for the person now holding it, and that person is now holding it alone.
The paper step also has a limit of its own. Once the report leaves the counter, nobody sees what was read, what was understood, or what the patient did next. The institution’s records show that a report was produced. They show nothing about the question the report left behind. A step that ends on paper cannot be measured, which is why it keeps producing the same fault quietly.
How medical oversight attaches to the reading of that report is a separate subject, documented from start to finish in the article on human oversight in AI decisions. The claim kept here is narrower. The report cannot be read, a question arises because of that, and the moment it does it leaves the business.
The question goes to an unvetted source and the business never sees it
The first route is easy to follow. The patient looks at the sheet, snags on an abbreviation, opens a phone and types that abbreviation into a search box. The question leaves the hand. Seconds later it arrives: at whatever unvetted sources the search results happen to surface that day.
There is a real answer at the end of that route, which is why the fault is easy to underestimate. The problem is not an absence of answers. The problem is that nobody can say who produced this one, what it rests on, or who to go back to if it turns out to be wrong. The patient takes it home and acts on it. A decision now exists with no author behind it.
Seen from the business, the entire route is dark. Which value was queried does not appear in your records. Which abbreviation caused the confusion does not appear in your records. How often the same question came up, which test type it clusters around, which group of patients gets stuck most often: none of it appears in your records. The question arose outside, was answered outside and closed outside.
So the answer to whose desk it ends on is short here. Nobody’s. Because no party takes responsibility for the answer, no correction is possible either. A patient who acted on wrong information turns up at a later appointment, and by then the origin of the error cannot be traced. Repair cannot start until the address has a name.
A short explanation takes up a slot in the appointment calendar
The second address is closer to the business and considerably more expensive. This time the question travels to the right place: the patient’s own doctor. But the only way to reach a doctor is through the appointment calendar, so a short explanation ends up consuming a slot that was reserved for something else.
What was spent, and what remains in exchange? What was spent is an appointment slot, the scarcest resource the institution has. What remains is an explanation that was spoken in a room and ended when the room emptied.
The doctor answers the same question again and again. That repetition is the signature of the fault. If the same explanation has to be given a second time out loud, the first one clearly did not stay anywhere.
A spoken answer lives exactly as long as the conversation. The patient carries it away, and it leaves no line in the patient’s file and no line in the institution’s records. The next patient snags on the same abbreviation and asks the same question from scratch. The only thing that accumulates is a count of how many times the explanation has been repeated, and nobody keeps that count.
In the first pattern the answer had no owner. Here it has a clear owner and a single use. The difference looks small and is not: an unowned answer erodes trust, while a single-use answer eats capacity.
An appointment is not always the wrong address. If the question is genuinely a diagnostic question, the doctor’s calendar is exactly where it belongs. What produces the fault is the same explanation being given out loud over and over and accumulating nowhere. A question that does not repeat falls outside this pattern entirely.
When buyer and seller meet through an acquaintance, the match leaves no record
Plastik ve Makina is a sector-specific classifieds platform where new and second-hand plastics processing machines are listed. Before it existed, the event ran like this. A manufacturer needs a machine and calls someone he knows. That contact points him to another. A few phone calls later a machine has been found. The match happens. The trade happens. No record remains.
The question had in fact been asked somewhere first. On general classifieds sites, machines like these sit as a single sub-line under a broad industrial heading. The buyer looked there, did not find what he needed, and went back to the phone. So a question was put to a system and the answer came back at the wrong scale. He was asking about the distinctions inside one machine family and got an entire industrial category in return.
The match then closes on a third person’s desk. The intermediary knows who is looking for what, knows who is selling what, and knows both parties. All of that knowledge sits with him. How many times a particular machine was asked about, which capacity buyers keep searching for and failing to find, appears on no list anywhere.
Payment and negotiation stay with the two parties, and no brokering layer sits between them. A trade closing outside the system is a deliberate scope decision. What produces the fault is the first question, the one asking who has this machine, getting answered without ever touching the system. If that question could be put to a catalog, at least the question itself would remain.
The seller has an unanswered question too: who is looking for this machine
Looking for the unanswered question only on the customer side is half the picture. The same gap has a supply-side face, and there the situation is worse.
A seller cannot put a machine in front of exactly the person who is looking for it. The listing goes up, lands under a broad heading, and the right buyer never opens that heading in the first place. Up to this point the description mirrors the buyer’s. The asymmetry starts one step later.
The buyer’s question was going to the wrong address. The seller’s question has no address at all. Who is looking for this machine cannot be typed into a search box. It cannot be put to a list. It cannot be turned into a filter. Because demand is collected nowhere, the seller is left with guesswork: drop the price, refresh the listing, call a few people and see.
Both sides are staring into the same gap, and the gap is what keeps them from seeing each other. The buyer cannot see the machines on the market in one place. The seller cannot get a listing to the right person. Whether they meet at all is decided somewhere neither of them can observe.
How that gap gets translated into a catalog decision, meaning which distinctions the category axis is built on, is not this article’s subject. That decision is written up with the machine family example in the article on modelling product variants. What stays here is the diagnosis alone: the unanswered question has two sides.
The question that is never asked ends the sale quietly
The fourth pattern is the quietest, because in this one the question is never asked at all.
Kokudankokuya is a perfume store, and its customers decide without smelling anything. A gesture that takes seconds in a physical shop becomes a gap on screen with nothing standing in for it. The customer sees the name of the fragrance, the bottle size and the price. What the customer cannot see is whether the scent suits her.
A second layer sits on top of that. Part of the shopping is not done for the buyer at all, but as a gift. The person deciding is not the person who will wear the perfume. The question being carried is not about her own taste but about someone else’s, and finding a place to ask that is harder still than asking about your own.
Why the question reaches no channel becomes visible in the record model. The catalog record holds the fragrance name, the bottle size, the price and the stock. Those are the fields a sale needs and every one of them is correct. There is simply no field where will this suit me could be written. A question with no counterpart in the record cannot leave a trace anywhere, no matter how many channels the business opens.
What remains is nothing. No message, no call, no search log, not even an abandoned cart line. Only a decision that was never made. Of the four patterns, this is the one a business notices last, because there is no event to notice.
The edge of this pattern is clear as well. A customer who knows a fragrance by name and reorders the same bottle has no unanswered question at all; the catalog answers her by name. The pattern appears with a first-time buyer, and most sharply with someone buying for another person. How a browsing axis built on the character of a scent gets constructed is a separate subject, and the mechanism sits in the variant modelling article.
The four patterns side by side
Four patterns converge on a single measure. The question leaves the customer, arrives at an address, and that address is not where the business keeps records. The addresses differ, the shape of the fault differs, the result does not.
| Fault pattern | Where the question goes | Whose desk it ends on |
|---|---|---|
| A report that takes medical training to read is handed to the patient | To unvetted sources in search results | Nobody’s; an answer is produced with no party behind it |
| The same explanation is requested out loud again and again | Into the specialist’s appointment calendar | The doctor’s desk; the answer stays spoken and never reaches the next patient |
| Buyer and seller cannot find each other on a sector-specific list | To the phone and the chain of acquaintances | The intermediary’s desk; the match leaves no trace on the platform |
| The customer decides without trying the product, and often for someone else | Nowhere; the question is never voiced | No desk at all; what remains is a decision that was never made |
Working out which row runs in your own business is the first step of any repair, because the addresses call for different repairs. A question travelling to an unvetted source and a question that is never voiced do not get fixed by the same move.
How two of these four faults were closed is documented elsewhere: the distinctions a catalog axis is built on in the product variant article, and the way medical oversight attaches to a readable text in the human oversight article.
Underneath all four patterns lies the record itself. If a question has no counterpart in the record, no answer forms there either. Which fields get kept, which question gets bound to a field, and which record carries which question are decisions taken while a system is being set up rather than after it runs. The web development service page describes where those decisions get made. If you can already name the pattern running in your own business, that is the decision waiting next.
