Back to blog Connectivity

A2P SMS Encoding Integrity: Testing and Troubleshooting Guide

Learn how to run controlled text checks for A2P SMS, compare the available evidence, and clearly report what cannot be verified.

Diagram of an A2P SMS flow with checkpoints for comparing text and available records

What preserving encoding means

In this guide, text integrity means comparing the intended content with the copies and records available during sending. It does not mean that every part of the route is visible or that the text shown in a record is necessarily an exact copy of what the handset received.

It helps to separate three questions: What text did the application build? What content was sent to the interface? What evidence is available about its processing or status? A difference between two observations helps narrow down what to investigate; on its own, it does not prove which component caused it.

  • Record the input text and each available observation separately.
  • Distinguish an observable discrepancy from a confirmed cause.
  • Treat content comparison and reported status as separate checks.
What preserving encoding means

Observation points: from the application to the destination

Map the actual route used by your integration before running checks. Include the application and its templates, the interface you use, the components under your control, and any observation points the provider can make available. Include the destination as a stage in the route, without assuming you will have access to the internal contents of other systems.

Keep a copy of the text before sending it and record the information returned by each component. Implementations and available fields can vary; do not assume that a simplified record describes every transformation or lets you reconstruct the full route.

  • Note the application version, components, and configuration used for each check.
  • Keep the identifier returned by the interface and link it to later records when available.
  • Ask the provider what each record represents and what its limitations are.
Observation points: from the application to the destination

Characters and representations: what to record

Text may be represented or handled differently by components in an integration. The evidence available here does not establish which representation a particular route uses or how each component handles it. For that reason, do not infer those details from how the text looks in a console alone.

Prepare a character list based on the content you actually need to send. Record the exact text before sending and any processing information the interface exposes and documents. Do not assume that two systems calculate length or describe content in the same way.

  • Save a copy of the intended text before sending.
  • Note any changes the application makes before calling the interface.
  • Record additional data only when it is available and documented for that integration.
  • Do not treat a counter displayed by one system as proof of how another system represented the text.

Design authorized, controlled checks

Use only test destinations and senders you are authorized to use. Keep the configuration and text constant when comparing results. Change one variable at a time to make discrepancies easier to narrow down.

Prepare small, reproducible test cases with characters from the content you actually plan to use. Repeat a check if the result varies, and note the time, test destination, sender, identifier, and configuration. The available information does not support setting specific technical criteria for a particular representation or length.

  • Confirm in advance that you are authorized to use the test destination and sender.
  • Send each case separately and use a test label that contains no personal data.
  • Compare configurations only after you have established a stable baseline.
  • Do not use checks to contact unauthorized recipients or bypass controls.

Compare evidence without collecting unnecessary data

Compare the intended text with each copy available in your system and in provider records. If the provider does not expose the processed content, record that limitation and request only the minimum information needed to narrow down the issue. Do not infer content from an acknowledgment or reported status.

To limit exposure, use synthetic messages and authorized test destinations. Restrict access to records and avoid saving the full message body when a case reference or controlled comparison is sufficient. If you need to retain text to reproduce a discrepancy, define access and retention periods in line with your applicable policies and obligations.

  • Keep a table with the stage, observed value, source, time, and limitations.
  • Distinguish data from your system from information reported by the provider.
  • Do not include real one-time passcodes, sensitive information, or personal data in checks.
  • Keep enough evidence to reproduce the case, not an indiscriminate copy of traffic.

Interpret discrepancies without assigning causes too soon

A character that appears different, is missing, or has been replaced is a signal to investigate, not automatic proof of cause. First check the text generated by the application; then compare the successive observations available. If you know only the input text and a final status, the responsible stage may remain undetermined.

If the discrepancy coincides with other observed changes, record them without assuming a causal link. Differences between the data held by two systems do not, by themselves, prove what happened in the intervening stages or what appeared on the handset.

  • Reproduce the case with a minimal message containing the problematic text.
  • Compare it with a control version and document the observations for both.
  • Check whether the application changed the text before attributing the issue to another stage.
  • Mark any stage without direct evidence as unconfirmed.

What acknowledgments, records, and DLRs do—and do not—prove

Interpret an acknowledgment or DLR according to the definition provided by its source. The available evidence does not establish a specific meaning for a status from an interface or provider. Do not automatically present a reported status as proof that the user read the message or that the text reached the handset screen intact.

Before using a status as an operational criterion, ask for the code definition and known limitations. If there is no independent check of the handset, describe the result as a reported status, not as verified receipt on the device.

  • Document the meaning of statuses according to the source that reports them.
  • Distinguish a request acknowledgment, a reported status, and an independent check of the handset.
  • Do not use a DLR as proof of reading or as confirmation of text integrity.

Troubleshooting procedure and acceptance criteria

When you find a difference, keep the smallest reproducible case and note the configuration. First review the text generated by the application; then compare what was sent to the interface with the response received; finally, ask the provider for the available records associated with the relevant identifier. When escalating, include the time, authorized destination, sender, test text, and exact observations. Ask the provider to clarify what stage each record represents.

Define criteria based only on what you can observe and compare. For example, you can require that text match at accessible points and that discrepancies can be reproduced or narrowed down. If you cannot verify content at a particular stage, state that limit and agree on how incidents will be handled, rather than turning an assumption into an acceptance criterion.

  • Establish a baseline using synthetic messages and authorized senders and destinations.
  • Assess the route against verifiable requirements, not assumptions about invisible stages.
  • Ask the provider for definitions of records and statuses, as well as any documented information that affects your integration.
  • Repeat checks after changes that could affect the text or available records.
FAQ

Frequently asked questions

Does a DLR confirm that the text was received intact on the mobile phone?

You cannot conclude that from a DLR alone. Interpret it according to the definition provided by its source; by itself, it does not prove what text appeared on the handset or that the user read it.

How can I find out how text was handled in my integration?

Compare the intended text with the observations available at each stage and consult the provider’s documentation. How text looks in a console is not enough to confirm its internal handling.

What cases should I include in checks?

Start with the actual content you need to send and test small, reproducible variations. Use only authorized destinations and senders.

Does a difference between the original text and a record prove that the provider changed it?

No. It shows a discrepancy between two observations. To narrow it down, compare the successive evidence and confirm what each record represents and what information is unavailable.

What should I send the provider when escalating an issue?

A minimal reproducible case with the time, identifier, configuration, authorized destination, sender, and an exact comparison between the intended text and available observations. Avoid including personal data or real sensitive messages.

Sources consulted

  1. 3GPP specifications3GPP
  2. ITU-T E.164International Telecommunication Union
  3. GSMA resourcesGSMA
  4. SMPP specificationSMPP Developers Forum
  5. HTTP semanticsIETF