Do public "where to find orders in Telegram" roundups work

What we checked

Five search queries returned seventeen articles that present themselves as roundups of places where a contractor can look for orders. Telegram links turned up in ten of them. From those links we dropped service paths and every name ending in "bot," which left 216 unique addresses.

One source is counted separately, and that needs to be said plainly. The largest roundup is titled as a list of channels with job openings, and it contributes 98 addresses. Counting it on a par with the others would mean proving something already known in advance, so the main sample excludes it and the full one is shown alongside. Do not derive one from the other by subtraction: the sources overlap.

For every address we asked two questions: does it exist at all, and what is it, a channel, a group chat or a personal account. The type was determined from the counter line on the address card: "subscribers" versus "members."

How many addresses turned out to be live

Of the 216 addresses, eleven do not exist: Telegram returns a service placeholder for them. That is a noticeable share for lists that are published as ready to use. Roughly every twentieth address leads nowhere.

Live addresses number 205. Of these, 161 are channels, 41 are group chats and three could not be identified. In other words, about eight in ten of the recommended addresses are a channel rather than a chat. A caveat is needed here, and we make it deliberately: channels sometimes have comments through a linked group, and we did not check that. So "you cannot reply there" does not follow from our data. What does follow is something else: the main flow in such an address is one-way, and by default there is no provision for a conversation with the author of a request.

In the main sample without the job-openings source the picture is the same: 135 addresses, six nonexistent, 109 channels and 17 groups.

What gets published inside channels

A public preview was available for 159 of the 161 channels. The side of the market was determined by the frequency of non-overlapping speech markers in the text of the messages themselves, not by the name of the channel. Names lie systematically, and "orders for a designer" regularly turns out to be a feed of staff job openings.

On the hiring side we counted phrases such as "vacancy," "resume," "experience required" and "we are hiring full-time." On the order side we counted "can anyone recommend," "who can do this," "I will order," "who will take this on" and "how much will it cost." Words that occur on both sides were not counted at all, and words about the work format, such as "remote," are tracked separately and do not affect the conclusion.

The result: 118 channels are job feeds, the number of channels on the client side is zero, and the classifier could not sort another 41. About those 41 we claim nothing: more than half of them are simply empty, with fewer than six messages in their whole history.

Why this is not a quarrel over words

A job opening and an order lead to different scenarios. In a job feed a person becomes an applicant: they send an application, wait, compete with a hundred others just like them and at best end up with a fixed pay rate. In a client's message the same person is a contractor: the client has described a problem, and the conversation is about the task, not about a resume.

An article titled "where to find orders" whose content is job feeds sends the reader somewhere other than where they were going. They do not get a refusal. They get a different kind of work.

How to check a chat yourself in a quarter of an hour

Read the last hundred messages without joining, if a preview is available. Count two piles: how many messages offer services and how many ask about a task. If there are three times as many offers, it is a chat of colleagues, not of clients.

Next, look at whether questions get answered at all. A community where a question hangs unanswered for a day is dead, regardless of the number of members. And read the pinned message: in many chats any offer of services means a ban, and the rules say so outright.

The sign of a good chat is counterintuitive: few of your colleagues are in it. A community of plumbers is useless to a plumber and very useful to someone who sells tools to plumbers.

Limits of the measurement

Group chats cannot be read from the outside, because groups have no public preview. We did not join them and do not judge their content. Judging what you have not seen is exactly the mistake that made us run this measurement in the first place. Two of the 161 channels returned no preview and were left out of the analysis.

The measurement data is open for citation under the Creative Commons Attribution 4.0 license, with the source credited. We do not publish the addresses from other people's lists as a list of our own: the point of the measurement is not to replace one unverified list with another.

What the measurement taught us

The first three runs produced a confident wrong answer, and none of them looked like an error: the code ran without exceptions, the output was plausible and the server answered with code 200.

It turned out that a nonexistent Telegram address does not return an error but an ordinary page with a service title. The check "no title means no such address" did not catch it, and the first run declared every address live, including three made-up ones that we mixed in as a control. Without those three names we would have noticed nothing: a hundred percent of addresses live looks like a successful list, not like a broken check.

The second run tripped over something else: under frequent requests Telegram answers with the same code 200 but returns an empty page. From the outside this is indistinguishable from a nonexistent address, and one hundred and forty-five live channels suddenly became "accounts." The fix is a retry after a pause, not a conclusion.

From this comes a rule that is worth more than the numbers themselves: an infrastructure failure has no right to turn into a statement about the data. We now have three outcomes instead of two: "exists," "does not exist" and "could not find out," and the third does not go into the statistics at all.

How to check other people's lists in general

The first thing to look at in any article with a list of communities is the date and a sign of verification. If it does not say when the author opened these addresses, most likely the author never opened them. A list gets copied from article to article for years, and dead addresses travel along with live ones.

The second is to tell a channel from a chat. This is visible in a second from the address card: a channel shows the number of subscribers, while a group shows the number of members and how many of them are online right now. In a "where to find orders" roundup, a channel almost always means a job feed.

The third is to read not the description of a community but its messages. The description is written by the owner, and it tells you how the community was meant to be. The messages show what it has become.