Table of Contents
Some unsolicited career advice: When the platform owner invites you to partner with them, say yes. In telecom, if an operator asks for your help, say yes. In fact, your yes should come before the inflection in their question even fades. That’s what I did when AT&T’s Alex Bobotek called me in early 2017 to join a project that would eventually become 10DLC.
I was in my Santa Monica office, with its peekaboo view of the Pacific, when Alex called. AT&T was collaborating with other partners on a message-tagging pilot. The goal was to see if “tagging” a text message could make the sender’s identity clear. While the near-term aim was to bring end-to-end message traceability and shortcode-like predictability to long code deliverability, the ultimate goal was to identify who was sending the message.
Sidebar: The State of Texting Circa 2012–2018
By 2015—the year the Apple Watch debuted—long-code texting was booming. Innovative products introduced by A2P providers like Twilio, Bandwidth, and a growing ecosystem capitalizing on the medium’s potential drove this surge. The insatiable demand for businesses to connect with their customers fueled massive growth. The problem? None of this growth was carrier sanctioned. While there were select enterprise-to-person (E2P) use cases and allowlisted numbers, there wasn’t a clearly established long-code “product” enabling applications to text consumers.
For a detailed breakdown of long-code texting, read my unpacking of Twilio’s Zipwhip acquisition.
The Challenge: Differentiating A2P from P2P
For any off-net text messaging—such as when a message went from an AT&T subscriber to a Verizon subscriber—it was routed through an inter-carrier vendor (ICV), either Syniverse or SAP (now part of Sinch). Carriers assumed that every mobile-terminated (MT) text coming from an ICV originated from another subscriber on a different network and would let it pass through. When spam entered this P2P channel, they treated it like a coordinated SIM farm attack, employing volumetric controls and content moderation, particularly for unknown links.
The fundamental challenge was their inability to differentiate whether a person or an application sent the message. In a classic case of adversarial interoperability, the market had created a pathway for A2P texting, yet the technical infrastructure for differentiating A2P from P2P simply didn’t exist. This gap between P2P and A2P traffic handling set the stage for the tagging pilot. One that aimed to redefine how carriers and applications managed long-code messaging.
The Collaboration
The pilot brought together AdaptiveMobile, Bandwidth, EZ Texting, Syniverse, and, of course, AT&T. As the firewall vendor, AdaptiveMobile detected spam across networks worldwide. It was important for it to categorize A2P vs. P2P messages. Bandwidth and EZ Texting were the OTT providers (a term with a different meaning back then!), while Syniverse served as the gateway through which messages entered the AT&T network. Each organization was responsible for securing its own management and legal approvals. From the outset, Alex made it clear: This was a technical pilot focused solely on testing a technical solution. Pricing, onboarding, and other product-related discussions were explicitly out of scope.
The Technical Approach
The first challenge was finding the most efficient way to determine whether a text message was A2P or P2P. Allowlisting phone numbers wasn’t a viable solution. CPaaS providers frequently purchase numbers in bulk—often tens of thousands—and reassign them between customers, making it nearly impossible to identify the true sender. Content analysis was also ruled out due to the sheer diversity of use cases. This left one option: tagging each individual message to reveal its true nature.
Could the SMPP specification used to transmit text messages be extended? Since version 3.4, the spec has included the ability to add optional fields through the tag-length-value (TLV) encoding scheme. This made it possible to introduce fields to designate a sender ID, a campaign ID, or even a throughput metric.Think of TLV encoding as a flexible way to add labels to a message, like tagging fields in a database. However, implementing this was far from simple. Every aspect of the messaging infrastructure and every system handoff had to be rigorously tested to ensure nothing broke.
At one point, Syniverse brought in Chris Wright, one of the original authors of the SMPP specification, to confirm that using an unused segment of the message header (source_subaddress) would not compromise the network.
For the next year, the teams met monthly—sometimes more—to test and review. No product pitches and no legal fluff—just engineers talking to engineers. By early 2018, we had a working demo. The message header could, in fact, distinguish A2P from P2P traffic in a meaningful way.
EZ Texting even sponsored a tech workshop at Mobile World Congress that year that was attended, in person and virtually by all the carriers, and other players of the ecosystem. In a packed beach-side hotel conference room, a team of product and technical experts spent an entire day discussing the technical details of the implementation and the next steps.
EZ Texting Messaging Workshop, September 10, 2018, Santa Monica, CA1
The Unforgiving Small Business Customer
The SMB is loyal and unforgiving. Deliver the value you promise at a fair price, and they’ll stick with you as long as they’re in business. But the moment service slips, they’ll bail—especially in software, where competitors are just a click away. In this sense, an SMB software provider is more like a neighborhood diner than a national chain. If Taco Bell serves you a bad meal, you might blame the location or chalk it up to an off day. Rarely would you think the Taco Bell brand itself is lousy. However, if a local restaurant serves you a bad meal, you’re probably never going back. The same principle applies to SMBs and messaging.
At the time, EZ Texting was focused on short codes. It dabbled in long-code texting and, along with Zipwhip, was one of the few providers enabling landlines for texting nationwide. But its core business was shortcode messaging and voice transmission. Gray routes—where every new customer required special downstream arrangements—weren’t viable. The risks of high customer attrition were simply too great. Yet market trends were clear: More businesses wanted to text from long codes.
Participating in the pilot gave EZ Texting the opportunity to not only partner with the larger ecosystem, it was also the way to give feedback on what was important to the SMB. All they cared about was for long-code texting to work “just like” it does on their cell phone.
Why “A” and not “The” Is in the Title
When I share this story with others, I am told that this idea of 10DLC as a carrier-sanctioned route has been attempted before. This pilot, for example, was built on a project from 2012, when Syniverse and Grand Central (later Google Voice) worked to properly tag landline bridge numbers. This is not surprising. In a network as complex and historic as telecom, it would be haughty to believe this was the first attempt to do something. It’s been true for other technologies, and it’s true for messaging. Toll-free texting, for example, had been around for a while (most famously, Heywire), but Zipwhip managed to scale it. What made this attempt at 10DLC stick was the carrier’s sponsorship and an expansive use case.
What The Trial Didn’t Address
What we proved to the industry was that it was possible to differentiate between A2P and P2P traffic. We demonstrated that a message payload could include a campaign ID, sender ID, and throughput parameters. However, we didn’t define what those parameters should look like or who would decide them—that work was left for product development.
The pilot ended in 2018, and over the next six years, more carriers joined the effort (though I believe AT&T is the only one still using message tagging). The challenging task of identifying and scrutinizing senders at scale gradually came together. Today, 10DLC is a defined, sanctioned, and reliable system for A2P messaging, even if it can still be confusing at times.
Lessons Learned
For me, there were four key lessons the tagging pilot taught or reinforced.
- Partnerships thrive on trust. High trust creates the space for curiosity and creativity, enabling teams to explore, experiment, and solve problems in unconventional ways. Without trust, the kind of collaboration we had during the pilot simply wouldn’t have been possible. In fact, I don’t think we even signed NDAs!
- The best partnerships have a ruthless alignment of outcomes. When everyone is rowing in the same direction, the work feels inevitable—almost vicious in its focus. I don’t think any of us doubted we’d figure it out. There was a certainty to the outcome because we all shared the same goal. When EZ Texting joined the pilot, it wasn’t about competing or proving we were the best; it was about contributing to a shared vision. The alignment came from knowing that if long-code texting didn’t mirror texting from a cell phone, the SMB customers we represented wouldn’t adopt it.
- The network is a palimpsest. It’s a layered, ever-evolving system where new ideas and technologies are built on top of what came before. This pilot itself stood on the shoulders of earlier work done, understanding that made the chances of success even more likely..
- It’s always about the people. It is the relationships you build, the trust you foster, and the teams you create—regardless of employer—that determine your success.
Finally
Alex called me because I had been voicing frustrations about the inaccessibility of long-code texting for my customers for years. So when he was ready to try out a solution, he asked if EZ Texting wanted to be part of the trial. What he was really asking was if I was ready to stop whining and help fix it.
And that’s the final lesson: trust, alignment, history, and people. Everyone trusted each other and the process. Together, we built the first mile marker for 10DLC.
1 In the picture (top to bottom, left to right): Participant 1, Participant 2, Stephanie Lashley, Simeon Coney, David Diggs, Mike Reading, Obi Hendrickson, Brad Roldan, Brad Blanken, Alex Bobotek, TJ Thinakaran, Kyle Spinks, Radu Maeirean, Lori Leong, Ian Matthews, Participant 3, Jason Sommerset, Shannon Donohue, Jaclyn Abrams, Eileen Noriega, Stefan Heller. Virtual: Kris Wetterings, Kara Lihosit, Susan Marion, Chris Wright (and others).