MVNOs, it’s time to Moneyball with AI

Editor’s note: “Que sera, sera”—a Spanish phrase that translates to “Whatever will be, will be”—is a reminder of life’s unpredictability and the need to adapt with resilience. This week, I was supposed to be in Valencia, Spain, rocking the stage with the keynote at the MVNO Nation Live event. Instead, my left kidney decided it had other plans. So, no Spanish soil or paella for me—WAH! Instead, I landed in a hospital dealing with a rogue kidney stone, a kidney infection, and an unscheduled surgery. The scenario screams “Que sera, sera” alright, making it crystal clear (no pun intended) that life’s got its own script, and I just have to roll with the (kidney) punches (and apparently drink A LOT more water on the regular). A huge shout-out to the MVNO Nation Live crew and attendees—I’m so bummed to have missed the event!

DR Signature

MVNOs, it’s time to Moneyball with AI

Remember the 2002 Oakland Athletics, a baseball team in the United States? If you’re not familiar with the story, let me set it up for you. This team had one of the lowest payrolls in Major League Baseball. The squad was full of rookies, players who had been written off, and veterans at the end of their careers. But General Manager Billy Beane had a different idea. Instead of adopting the same old strategy everyone else used to chase star players (which he couldn’t afford), he flipped the script. Instead, he used DATA to find value that nobody else saw—and he changed the game forever. Michael Lewis chronicled the team’s season in his book Moneyball: The Art of Winning an Unfair Game, which was made into a great movie by the same name starring Brad Pitt and Jonah Hill.

MVNOs have the same opportunity in the world of telco today. As an MVNO, your network provider’s legacy, on-premise tech stack has kept you from getting killer ideas to market quickly, and slowed your ability to attract and engage subscribers. Armed with the power of the public cloud, artificial intelligence (AI), and your entrepreneurial spirit, you can turn the tables—and beat MNOs. Instead of just offering services, you can craft experiences. Instead of retaining subscribers, you can make them your biggest fans—all while growing your revenue at the same time. The fact that you don’t have the burden of managing the network and all that legacy tech means you can use your time to experiment and change the game.

To do this, you have to be like Billy Beane: you have to think differently, cozy up to the public cloud, and embrace the real magic: using AI—specifically machine learning (ML)—to design plans tailored to individual subscribers.

Totogi’s Charging-as-a-Service and Plan AI

I’m talking about Charging-as-a-Service and Plan AI products from Totogi, where I’m acting CEO. These world-class, carrier-grade, software-as-a-service (SaaS) products learn from subscribers’ choices and use that knowledge to enhance the experience for everybody. Sure, everyone is familiar with our charging platform. But the mistake people make is that they think it’s a plain, old, regular charger—like the ones out there from the dead and dying charging vendors. Totogi is different: we aren’t just recording transactions in a database, where they’re waiting to be sent down the line to a billing system. Instead, we are collecting highly-structured charging data. When you apply ML models to it, it’s filled with rich insights you can use to find new subscribers, stop churn, and grow ARPU.

Totogi aims to build telco’s BEST ML model. FULL STOP. That’s what we say everyday in our company. Where’s the bar? What’s the expectation? What’s the definition of great? And we won’t stop until it’s the world’s best. We feel there is power and differentiation in investing in ONE model that the industry uses. We can pour millions of dollars into talent, compute power, specialized chips and data to build something for MVNOs that they would never, ever be able to afford on their own. And this is THE reason you buy Totogi: for revenue optimization, not just transaction processing.

Compare this approach to other vendors that say they are adding AI and ML to their charging products. Just ask them one simple question: “Is this my own ML model, or is it shared?” I bet you $10 that the answer will come back as, “Of course this is your own model customized for you, running on your own data.” And that approach is DEAD WRONG. The custom model (one model per customer) is laden with hidden costs, operational missteps, and a perpetual game of catch-up with evolving ML technologies—never mind not having nearly enough data to train the damn thing. It’s expensive and won’t work, especially for MVNOs.

Totogi’s One Model Approach

IT’S ALL ABOUT THE DATA. Totogi has taken a one-data model because we feel it’s a superior approach. Do you know how much data it takes to train an ML model? You need not only a quantity of data, you need diverse data. Training an ML model on only your own data runs the risk of “overfitting,” where you get too many false positives because it’s trained on data that’s too similar (all yours). On top of that, you need well-structured data. What I mean by that is data that is all in the same format. Do you think Amdocs, with its penchant for customizations, can do this with its customers? Plus, its customers’ data is spread all over the place: some on premise and some in the cloud, in different schemas, and on different platforms and versions. Good luck normalizing that.

CHIPS AND MORE CHIPS. Totogi has famously gone ALL-IN on the public cloud. Some people still think I’m wrong with this call. But every year, it becomes more clear: when you’re on the public cloud, it’s EASY to use other things in the public cloud, like the AI chips you need to bring costs down while training a ML model. Thanks to AWS’ investment in Trainium and Inferentia, Totogi gets to use those chips whenever we need to. How will legacy on-prem vendors do this? Are you, as an MVNO, going to buy a set of sold-out $100K Nvidia servers to train your own models? Are you going to move your data to the cloud to get access to the AI chips, like Totogi does? Again, I think other vendors will struggle here.

CONSTANT IMPROVEMENT. ML models need constant maintenance. You need trained data scientists (which in Silicon Valley will run a cool $600K – $1M in annual compensation) to tweak and update the models continually. Again, is the best way to build a best-in-class model to have your own custom model, or to have a vendor maintain a custom model for you? We don’t think so. Let us hire these $1M resources and pour effort into a single industry model. As we maintain and manage just one model, we can distribute the cost among our customer base. Having one model allows us to pour all our efforts into constantly improving the model, where ALL Totogi customers benefit. It’s 100% trained on real telco data, with the sole goal of optimizing revenue for MVNOs. This kind of focus matters.

Everyone’s Favorite Objection

Hopefully, I’ve convinced you that while you CAN build your own model, it will be inferior to the Totogi approach. I also trust that you’re convinced that your vendor’s solution of building a model for you on your own data is not the way to go. But that brings me to everyone’s favorite objection: if we’re all using Totogi’s model, where’s the differentiation? 

To differentiate from other Totogi customers, it’s going to come down to how you UTILIZE the insights from the model. Every MNO and MVNO will do it differently. If Totogi gives you churn predictions, will you use that information? How quickly, and in what way? Will humans intervene and go with their gut, overriding the AI recommendations? Will you use our API calls and put the results into your call center, or self-help AI chatbot? Totogi will give you great suggestions to drive revenue, but it will be up to the HUMANS to get it into the market. We are enabling you, but I’ve seen so many organizations hesitate, doubt, and just plain stop because they are afraid of what could happen. And I’ve seen other organizations run with it and see GREAT results. As they say, fortune favors the bold. This will be what drives differentiation. 

Time to Moneyball

Around the world, big operators are investing heavily in AI to revolutionize their businesses. The good news is that you don’t need big bucks for this bandwagon. MVNOs can apply AI to streamline plan creation and bolster marketing initiatives. And you can use ML to detect customer churn and untapped revenue opportunities. It’s time to Moneyball telco—and with the accessible compute, chips, and ML tools of the public cloud, along with Totogi, you can be Billy Beane, too.

Recent Posts

  1. The promise of AI is no UI
  2. Unlock the true value of your BSS with AI
  3. Telco’s shift to the public cloud gets real
  4. Mi Casa es su Casa (if you’re Verizon)
  5. Why AI’s future might depend on nuclear energy

Telco in 20 Podcast - Tune in

Get my FREE insider newsletter, delivered every two weeks, with curated content to help telco execs across the globe move to the public cloud.

"*" indicates required fields

This field is for validation purposes and should be left unchanged.


Ep 64 – The cloud-first mindset

Hawaiian MVNO Mobi built a public-cloud-native, mobile core on AWS. CEO Justen Burdette talks about why it’s so great to work with cloud-first solutions.


SUPERCHARGE your MVNO with personalization

By using highly cost-effective, carrier-grade, truly cloud-native software, MVNOs can gain real-time insights and deliver superior subscriber experiences with the help of AI and ML tools.


MVNOs can use AI too

Here are five ways operators and MVNOs of all sizes are leveraging AI for game-changing results. Don’t get left behind!