DEV Community

Cover image for When Roaming Says 'Connected' but Nothing Works: A Narrow Gap Between Sister Operators
fcn06
fcn06

Posted on

When Roaming Says 'Connected' but Nothing Works: A Narrow Gap Between Sister Operators

You land in another country. Your phone shows full bars and "5G". Nothing loads. Or calls fail while texts work.

Every roaming and customer-care team knows this case. This article is about that one problem, not about all of roaming. It is also honest about what operators already do well today.


1. What operators already see (and handle)

Let's be fair first. Roaming is not a blind spot in general.

  • Connection (attach) is visible. The home network knows if the phone registered abroad, on which partner network, and on which radio type.
  • Internet traffic is visible in the common setup. In most data roaming, traffic is routed back through the home network ("home-routed"). So the home operator sees the bytes, and also signs of trouble: slow responses, many retries, DNS failures.
  • Bill shock is mostly under control. In the EU, caps and alerts are standard. Many operators also block or limit special networks (ships, planes) by default.
  • Fraud has its own tools. Usage records for fraud detection (NRTRDE) and real-time checks on the home side cover the main cases.

So "we cannot see anything" is wrong. The real question is smaller.


2. The one gap: "attached, but broken"

The home operator can often tell that something is wrong. It often cannot tell where.

A few typical cases:

  • Voice over LTE works at home but fails with one specific partner, because of a setting on one side.
  • A partner's local network is overloaded in one area, and sessions stall.
  • A routing or gateway problem at the partner drops some traffic, but not all.
  • The phone is fine, the home network is fine, and the fault sits in the visited network.

From the home side, these look the same: a customer with bars and no service. The home operator may then do the wrong thing: reset the session, tell the customer to restart the phone, or open a ticket that takes days.

What is missing is a quick, narrow answer to one question:

"Is this session failing because of something on your side right now?"


3. Why not just share the data?

The obvious idea is to send the visited network's logs and counters to the home operator. This usually does not work:

  1. Privacy and telecom secrecy rules. Continuously sending location and behaviour data about people across borders is hard to justify, even inside one company group.
  2. Cost for little value. Most roamers have no problem. Moving everything to help the few who do is wasteful.
  3. Different vendors, different formats. Every operator measures things differently.

So the data should stay where it is.


4. A simpler idea: send the question, not the data

Instead of moving data to the question, move a small question to the data.

  1. The home operator has a customer with a failing roaming session.
  2. It sends a short, fixed-format question to the visited operator: "Check this session. Is there a known problem on your side?"
  3. The visited operator checks its own live data, inside its own country.
  4. It sends back a short answer from a fixed list, such as local_congestion, known_fault, no_issue_found or cannot_tell.
  5. The home operator acts: tries another partner network, opens a ticket with the right team, or tells the customer it is not their phone.

If no answer arrives within a short time, the home operator does what it does today. Nothing gets worse.

Key properties:

  • Only a short answer leaves the country, never logs or raw counters.
  • The question is limited. It is one allowed question, about one session, not a free search.
  • No phone number or SIM ID in the question. A temporary reference (a pseudonym) is used instead.
  • Both sides keep a record of what was asked and answered.

A note on privacy: a pseudonym is not anonymous. Under GDPR it is still personal data. This design reduces exposure (fewer systems see the real ID, fewer logs contain it), but it does not remove the need for a legal basis and clear agreements.


5. Why would the other side answer?

This is the hardest question, and it is better to be direct about it.

Between competing operators, it is difficult. Sharing live network health can raise commercial and contract issues. It would need agreements, and probably an industry body or hub to be involved.

Between sister operators in the same group, it is much easier:

  • they share governance and customer-experience goals;
  • they often already have group-level contracts and shared teams;
  • group management can set the rules and the targets;
  • a small pilot between two countries is realistic.

That is the credible starting point. A group with operations in several countries, where customers travel between them every week, can test this without waiting for the whole industry.


6. What it would and would not fix

Likely to help:

  • faster, correct handling of "full bars, no service" cases between group partners;
  • fewer pointless resets and fewer long, back-and-forth tickets;
  • clearer evidence for who owns the fault.

Will not help (or not much):

  • problems caused by the phone, the eSIM profile or the customer's settings;
  • fraud, which is better handled with existing tools on the home side;
  • bill shock, which is mostly a policy-setting question.

Open questions a real pilot must answer:

  • Which few answer codes are actually useful to customer-care teams?
  • How often is the visited network really the cause?
  • What is the legal basis, and what do the roaming agreements allow?
  • Is the saving in handling time and tickets worth the integration effort?

7. The trust problem

Even a tiny query channel between two companies raises a practical security question: how do you let one company's automated system ask another company a question, without static passwords, without letting it ask anything it likes, and with a record both sides can check?

Normal encrypted connections protect the pipe. They do not limit what may be asked, and they do not give both sides a shared, checkable receipt.

In the next part, I will show how I approached this in my open-source project, Trust Gateway: one narrowly defined question, a limited permission to ask it, and a signed receipt of question and answer.


If you work in roaming, wholesale or customer care: how often is the root cause on the partner's side, and how do you find out today? I would like to hear where this is wrong.

Top comments (0)