Travel eSIMs are taking a larger share of the retail relationship. Seven failures around IR.21 data can quietly erode the wholesale roaming value operators still control. Roaming has a new revenue problem.

 

For years, operators worried about fraud, settlement disputes, steering, wholesale rates and bill shock. Those problems have not disappeared. But another pressure has become impossible to ignore: travellers increasingly buy their connectivity somewhere else.

Ahead of 2025, Juniper Research estimated that travel eSIMs could displace $1.2 billion in potential operator roaming revenue during the year. Its October 2025 market update subsequently estimated travel-eSIM package revenue at $1.8 billion, up 85% from 2024.

Yet the story is not simply “travel eSIMs versus mobile operators.”

Those eSIMs still need mobile networks underneath them. Kaleido Intelligence forecasts that travel-eSIM traffic could contribute close to $10 billion in attributable wholesale roaming revenue for operators by 2030. In its 2026 survey of 68 Tier 1, Tier 2+ operators and selected MVNOs, 25% of respondents said they had already seen incremental inbound traffic and/or wholesale revenue through direct or indirect relationships with travel-eSIM and alternative roaming providers.

That creates an uncomfortable possibility.

An operator can lose roaming revenue twice.

First, it loses some of the retail relationship when the traveller buys connectivity from another provider.

READ MORE: Travel eSIMs Are Dismantling Roaming—So Telecoms Are Fighting Back

Then it can lose part of the wholesale opportunity that remains because the information used to configure roaming between networks is outdated, incomplete or implemented inconsistently.

This is where IR.21 stops looking like administrative plumbing and starts looking much more like revenue infrastructure.

What IR.21 actually is

IR.21 is a GSMA Permanent Reference Document used to structure the technical, network and operational information operators share to support interoperability between their networks.

It contains information such as TADIG identifiers, network identifiers, signalling and interconnect information, service capabilities, operator contacts, LTE-M and NB-IoT information and detailed data relevant to 5G standalone roaming.

IR.21 itself does not authenticate a subscriber, route their traffic or create a roaming invoice.

It provides reference information that operators use when configuring the systems that make those things possible.

READ MORE: The Intricacies of Roaming Billing and Settlement

RAEX is complementary. GSMA describes RAEX as the mechanism that organises IR.21 information into a standardised format that roaming partners can use more easily. Roaming hubs use the related IR.85 specification for additional hub-specific information.

That distinction matters because the real danger is not simply having “bad IR.21 data.”

It is drift between the current information supplied by a roaming partner and what the operator’s live network, security and operational systems are actually configured to use.

GSMA highlighted that risk when introducing the latest RAEX/IR.21 framework in 2026, warning that inaccurate information can lead to poor customer experience and missed revenue from roaming and interworking services.

Here are seven ways that gap can destroy value.

1. The inbound roamer who never becomes revenue

This is the simplest and potentially most expensive failure.

An inbound roamer represents potential wholesale revenue for the visited network. But before that traffic can be monetised, the subscriber needs to be able to use the network successfully.

Roaming configuration depends on multiple pieces of partner information: network identifiers, signalling information, service capabilities, addressing and other operational parameters.

Now imagine a partner changes something important.

The updated information is supplied. One system gets changed. Another does not. A configuration remains based on an older reference state.

The roaming agreement is still valid.

The network behaviour is not.

A subscriber may fail to obtain the expected service or may end up using another available network instead.

And this is what makes the loss difficult to see: if the session never becomes successful traffic, there may be no corresponding roaming revenue record to dispute later.

There is simply less traffic than there should have been.

Fraud creates an event.

A failed opportunity can create nothing.

2. The partner network changes before your roaming proposition does

The disappearance of 2G and 3G makes reference-data accuracy increasingly important.

As of January 2026, GSMA reported that 170 operators globally had no 2G service, with another 87 closures planned over the following decade. For 3G, 190 operators had already closed their networks and another 107 closures were planned.

At the same time, only 280 operators had inbound VoLTE roaming agreements.

That creates a messy transition.

A roaming proposition that worked last year may depend on a capability that no longer exists today.

A handset that once relied on circuit-switched fallback may lose that fallback route. An older IoT deployment may depend on a technology being retired. Voice behaviour can change even though data still appears to work normally.

READ MORE: Comfone Deploys 5G SA Roaming with BroadForward SEPP

This is precisely why the latest RAEX/IR.21 framework introduced new reporting around planned 2G and 3G switch-offs. GSMA says operators can use those reports to see when other networks intend to close legacy technologies and adapt their roaming propositions accordingly.

That is an important distinction.

A planned network shutdown is not an unexpected outage.

If the information existed but never became configuration, testing, device-policy or commercial action, the resulting disruption was preventable.

3. The 5G SA agreement that does not deliver 5G SA

5G standalone raises the operational stakes again.

Current IR.21 specifications contain dedicated information relevant to 5G SA roaming, including SEPP models, MCC/MNC associations, well-known FQDNs, SEPP FQDNs, IPX-provider information and N32 parameters.

SEPP—the Security Edge Protection Proxy—is central to protecting signalling between 5G networks across operator boundaries.

The important point is not that IR.21 somehow creates the N32 connection.

It doesn’t.

It supplies information that operators need when configuring and operating that interconnection.

And 5G SA roaming is only now entering genuine commercial territory. During a GSMA webinar in August 2026, Telefónica’s José Antonio Vivanco Martos said two operators had recently signed what he described as the industry’s first official commercial 5G standalone roaming agreement.

More will follow.

But a commercial agreement does not guarantee a functioning service.

READ MORE: Wholesale Roaming in the Age of Travel eSIM

If the operational environment does not reflect current partner information, the result can be failed interworking or a service experience that does not match what was commercially promised.

The operator may have negotiated the agreement, deployed the 5G core, contracted the IPX relationship and launched the proposition.

Yet revenue can still disappear near the end of the chain because the configuration is wrong.

That is a particularly expensive place to fail.

4. One IoT problem becomes a fleet problem

A traveller notices broken roaming.

A connected device may not.

That is why IoT deserves its own category.

IR.21 contains detailed information covering LTE-M and NB-IoT capabilities. Depending on the technology and configuration, that includes areas such as supported IP versions, control- and user-plane data transport, PSM, eDRX, SMS support and other network characteristics.

Those details matter because IoT deployments are unusually sensitive to differences between networks.

A tracker may cross five countries. A smart meter may remain deployed for a decade. A payment terminal, vehicle or industrial sensor can depend on behaviour that varies between networks.

GSMA’s Mobile IoT guidance warns that differences in network configurations, supported functionality, architectures and spectrum deployments can affect roaming capability, device behaviour and overall deployment economics.

Now multiply one mismatch by 10,000 devices.

The result does not necessarily look like 10,000 customer complaints.

It might look like missing telemetry, repeated reconnection attempts, unusual battery consumption, intermittent service or a fleet that mysteriously behaves differently in one country.

And IoT roaming is becoming financially significant.

Juniper Research forecasts that wholesale revenue from IoT roaming connections will approach $2 billion globally by 2028, representing roughly 10% of the wholesale roaming market it forecasts for that year.

At that scale, keeping partner capability data aligned with operational configuration becomes part of enterprise service assurance.

5. The identifier mismatch that surfaces later in settlement

There is an important misconception to avoid here:

IR.21 is not the TAP billing specification.

TAP is part of the wholesale roaming billing and settlement process. RAP supports the handling of errors around that exchange. GSMA is also promoting Billing and Charging Evolution, or BCE, as a more flexible approach that can support increasingly complex commercial models including IoT, 5G standalone and network slicing.

But IR.21 and settlement are not completely disconnected either.

TADIG codes appear in IR.21, and GSMA’s TADIG documentation shows that these identifiers are used across several roaming processes and specifications, including TAP, RAP and RAEX.

The revenue risk therefore is not:

“An old IR.21 file automatically produces a bad TAP file.”

That would be too simplistic.

The real risk is inconsistent operator and partner reference information across roaming operations, clearing, billing and related systems.

If one environment associates a partner, TADIG or MCC/MNC relationship differently from another, exceptions can follow.

Then somebody has to investigate.

Revenue recognition can slow down. Reconciliation becomes more manual. Disputes become harder to resolve.

Newer settlement technology can make the process more flexible.

It cannot make inconsistent source information correct.

6. The partner quietly stops preferring your network

Some losses never appear as a technical incident or billing dispute.

They appear later in traffic volumes.

If inbound roamers repeatedly experience registration problems, incomplete service availability or inconsistent quality, the partner sees that too.

Its roaming engineers troubleshoot the incidents.

Its wholesale team receives escalations.

Eventually, reliability becomes commercial information.

That matters because operators do not operate in a static bilateral world anymore. Travel-eSIM providers, roaming hubs, sponsored-roaming models and alternative roaming providers are creating new routes for wholesale traffic.

READ MORE: 6 Common eSIM Mistakes You Should Avoid

Kaleido’s 2026 operator research found that only 4% of respondents expected inbound roaming traffic to decline over the following three years. It also found that a quarter had already gained incremental inbound traffic and/or wholesale revenue through travel-eSIM or alternative-roaming-provider relationships.

There is therefore more traffic available to win.

But traffic does not have to disappear dramatically.

A partner can simply find another route that works better.

Repeated configuration friction gradually becomes a reason to send less traffic your way.

7. The incident nobody can reconstruct properly

The final failure is less immediate commercially, but increasingly important operationally.

Traceability.

When something goes wrong, can the operator establish:

• what information the roaming partner had supplied;
• when it changed;
• which teams received it;
• whether the required configuration was implemented;
• when that implementation happened;
• what was actually running when the incident occurred; and
• whether remediation was subsequently verified?

This should not be confused with regulatory certification.

IR.21 is not a cybersecurity regulation. NIS2 does not require an “IR.21 audit,” and having an IR.21 change log does not make an operator NIS2 compliant.

But NIS2 does require essential entities within its scope to implement appropriate and proportionate technical, operational and organisational cybersecurity risk-management measures. Providers of public electronic communications networks and publicly available electronic communications services are among the entity types covered, subject to the directive’s scope and classification provisions.

Article 21 addresses areas including incident handling, business continuity, supply-chain security and procedures for assessing whether cybersecurity measures are effective.

So traceable operational processes matter.

The same principle becomes particularly relevant around emergency calling.

GSMA warned in July 2026 that roamers can encounter technical issues with emergency calls during the transition from circuit-switched voice to IMS-based VoLTE and VoNR, especially as 2G and 3G fallback disappears.

Current GSMA roaming information also provides operators with visibility into emergency-calling capabilities across networks.

Nobody should describe that as “IR.21 compliance.”

But an operator that cannot reconstruct why a critical roaming configuration differed from current partner information has an operational-control problem.

Receiving an update is not the same as implementing it

The obvious answer to all this is automation.

But automation alone is not enough.

A mature IR.21 process should be able to:

• ingest current partner information through the appropriate RAEX processes;
• distinguish meaningful operational changes from routine updates;
• identify the network, security, billing or operational systems affected;
• route each change to the correct owner;
• track implementation rather than simply receipt;
• validate that the resulting configuration behaves as expected;
• compare relevant operational state with current reference information;
• make that information available during roaming troubleshooting; and
• retain a searchable history of changes, implementation and remediation.

The most important distinction is between processed and implemented.

An update arriving successfully in an inbox proves almost nothing.

Parsing an RAEX record proves slightly more.

The question that matters is:

Does the live service now reflect the partner’s current network state?

Not every operator needs to solve this in the same way.

A smaller MVNO or an operator heavily dependent on a roaming hub may sensibly rely on managed services rather than building another internal platform.

At the other end of the market, operators with large bilateral footprints, significant inbound roaming traffic, IoT customers and 5G SA ambitions need something more demanding: continuous assurance between partner reference information and operational reality.

That is a different problem from document management.

Conclusion: protect the roaming revenue you still control

The easy story is that travel eSIMs are taking roaming revenue from operators.

The more interesting story is that they are changing where roaming value sits.

Travel eSIM brands increasingly own the customer’s purchase decision. Yet much of the underlying connectivity still runs through operator networks, creating new wholesale traffic at the same time as traditional retail roaming comes under pressure.

And the wholesale opportunity is growing.

Juniper Research forecasts global wholesale roaming revenue increasing from $9 billion in 2024 to $20 billion by 2028, driven partly by 5G SA and IoT. Kaleido expects travel eSIMs themselves to become a major contributor to operator wholesale revenues over the rest of the decade.

That changes the strategic importance of IR.21.

It should not be treated as a document operators periodically receive and file away.

Nor should it be exaggerated into something it is not.

IR.21 is reference information. RAEX helps exchange it. TAP and BCE address different parts of the roaming stack. None of them individually guarantees that roaming works.

The commercial capability sits in connecting those layers correctly.

Operators cannot stop travellers comparing travel eSIM prices. They cannot dictate how quickly partners retire 3G, introduce VoLTE, launch LTE-M or move into 5G standalone.

But they can control how quickly their own systems respond when those realities change.

They can know whether an inbound roamer should have attached.

They can know whether an IoT fleet is being given the services expected.

They can know whether the technical assumptions behind a 5G SA relationship still match reality.

And they can know whether the information supplied by a partner actually made it into the network.

Losing part of the retail roaming market to stronger competition is a market challenge. Losing the wholesale traffic you already had the opportunity to carry because your configuration no longer matches current partner data is a preventable one.