Understanding and Eliminating A2P Message Filtering

This post summarizes key insights from a webinar on understanding and eliminating application-to-person (A2P) message filtering, focusing on the 10DLC channel. Click Here to watch the entire webinar!

Who Filters Messages?

Messages can be filtered at three levels:

1. Messaging Provider: These providers, like Signal House, have built-in content filtering systems that scan messages for spam, scams, or SHAFT content (sexual content, harassment, alcohol, firearms, tobacco). They use keyword analysis and sentiment analysis.
2. Direct Connect Aggregator (DCA): DCAs are the intermediaries between messaging providers and carriers. They handle larger message volumes and often use more sophisticated AI-driven language models to detect unwanted content.
3. Carriers: Carriers, especially the “Big 3” (AT&T, Verizon, T-Mobile), handle the largest message volumes and possess the most data for training their filtering models. This makes them the most significant source of message filtering.

Why Messages Get Filtered (Even When They Shouldn’t)

While some filtering is intentional (e.g., blocking explicit content), “unintended filtering” can occur when legitimate messages contain keywords or phrases similar to those used in spam or scam messages. For example, a car dealership’s legitimate message about a “free” offer and “low rates” could be flagged due to similarities with a scam message.

How to Minimize ‘Unintended Filtering’

• Avoid “Pushy” Language: Remove words and phrases commonly associated with scams (free, act now, while supplies last, you don’t want to miss out, excessive use of dollar signs or exclamation points, etc.)
• Split Messages: Split messages heavy with language that could be used in scam messages into multiple, shorter messages.
• Understand Keyword vs. Algorithm Filtering: Keyword filters are binary (a specific word triggers filtering), while algorithms analyze the entire message for patterns indicative of spam or scams. Ask your aggregator which method they use.
• Focus on Consent: Ensure you have explicit consent from recipients to receive messages. High opt-out rates and “report junk” clicks can trigger filtering even if your content is technically compliant.

Key Questions to Ask Your Aggregator:

• “Do you have content filtering in place? If so, what kind?” (Look for algorithmic filtering or robust KYC processes—keyword filtering is a red flag 🚩)
• “Has your company been subject to fines or penalties for content in the past?” (This can indicate stricter filtering.)
• “How many hops are there between a message being sent and the carriers?” (Fewer hops mean less filtering.)
• “What are the unacceptable use cases for your company?” (This helps you avoid inadvertently violating their policies.)
• “Do you actively monitor for content drift, and how?” (Drift content is when your messaging deviates from the registered use case, like moving from customer support to marketing.)

What Happens When Filtering is Too High?

Excessive filtering can lead to:

• Warnings: from your messaging provider
• DCA-related flags: These can limit or completely block your ability to send messages
• Carrier flags: These can also restrict or completely block messaging, sometimes even affecting your entire brand (EIN damage)

Proactive Strategies for Avoiding Filtering:

• Know Your Customer (KYC) with Your Aggregator: Establish a clear understanding of your business, messaging practices, and customer base with your aggregator. This can help them understand your messaging context and reduce the likelihood of false positives. Create a detailed use case and review sample messages with them.
• Address Opt-Out Rates: A high opt-out rate is a strong indicator of non-consensual messaging and can lead to filtering and flags.

Message Failure Reasons Beyond Filtering:

Messages can also fail due to:

• Technical Issues: Problems with connections between providers, DCAs, and carriers.
• Inactive Phone Numbers: Sending messages to numbers that are no longer in service.

Q&A Session

• Q: My deliverability could actually be higher on another provider just because of how strict their filtering is. Am I understanding this right?

  • A: Yes, absolutely. Deliverability can vary significantly between providers due to differences in their built-in content filtering strength and the number of “hops” in the message delivery process. Meaning, if you send the exact same messaging on Company A and Company B, your delivery rate could be significantly different if, for example, Company A has aggressive content filtering in place.

• Q: I want to create a use case with Twilio about my business. Can you walk me through how you would approach that?

  • A: Gather all relevant stakeholders (account executive, their manager, support team, TAM) and walk them through your business, customer base, messaging practices, and examples of past messages. Ask direct questions about their acceptable and unacceptable messaging categories, and work with them to iterate and improve your deliverability. Aim for >95% deliverability

• Q: My aggregator told me I had a carrier flag because of too many messages to out-of-service numbers and is trying to sell me an additional service. Is this legit?

  • A: Yes, sending messages to inactive numbers can trigger carrier flags as it suggests non-consensual messaging (it’s expected a very, very small minority will provide a fake number/enter it wrong. A high percentage of out-of-service numbers is indicative of a generic phone list, or random dialing). Whether the additional service is legitimate depends on its specifics. Contact me (will@signalhouse.io) for a review.

• Q: I was told the reason I had poor deliverability was because of opt-out rates. Does that make sense?

  • A: Yes, high opt-out rates strongly suggest non-consensual messaging and can lead to carrier filtering and flags (blocks on your messaging ability).

• Q: So if I swear it could lead to my message failing? Seriously?

  • A: Yes, some providers have keyword filters for swear words. Ask your aggregator about their policies.

• Q: Do you anticipate the new administration affecting any of these rules?

  • A: Changes are possible, though not necessarily drastic. Potential areas of change include regulations regarding foreign ownership of the Campaign Registry (TCR) and discussions surrounding 1st amendment concerns related to message filtering. I also anticipate the 10DLC process becoming stricter over time, regardless of the administration, with potential yearly reviews and more stringent vetting to combat abuse.

• Q: So filtering is one reason why my delivery rate is not perfect. But how else can messages fail? Is filtering the only way?

  • A: No. Messages can also fail due to technical issues (problems with connections) and inactive phone numbers/drift content.

• Q: Do the carriers associate a score to your campaign? Can I get filtered more because I was filtered at all?

  • A: Carriers likely use proprietary scoring systems for phone numbers and campaigns, considering factors like past filtering instances. While a direct “score” isn’t publicly available, past filtering can make future filtering more likely.

Want to learn more? Stay tuned for upcoming webinars on specific topics like campaign approvals, throughput, and creating a strong use case.