SMS in Logistics

Nexmo’s Adaptive Routing Finds Its Way To a U.S. Patent

Published December 15, 2016 by Glen Kunene

What began as an internal optimization to enable Nexmo to deliver better quality to its text messaging customers is now a patented innovation. Nexmo’s proprietary Adaptive Routing solution for high-quality global delivery of text messages has been issued a U.S. patent (U.S. Pat. No. 9,485,169). It’s the first patent for Nexmo.

Adaptive Routing is an algorithm-based mechanism that continuously monitors the available routes for delivering text messages and dynamically selects the best one based on their relative performance data.

Eric Nadalin, who as co-founder and CTO of Nexmo is responsible for much of Adaptive Routing’s development, recalled how the concept was born and what he sees as the next evolutionary step for the patented solution.

The Origin Story

Nexmo was founded six years ago on the concept of building a cloud communications company on the direct-to-carrier model. This model would eliminate the multiple hops that text messages commonly make among carrier networks before being delivered to their destinations. “We ended up building this massive carrier connectivity,” said Nadalin. “At a given time, you’d have 10 to 100 options to deliver a message.”

While this was a boon for developers who could now use APIs from the Nexmo platform to integrate streamlined messaging capabilities into their apps, it presented Nexmo with the challenge of scalability and quality. “We have all this inventory,” Nadalin recalls of the sudden abundance of routing options. “What’s the best?”

Message routing was (and remains) mostly a manual process, and deciding which was the best route for a given message was difficult without a reliable measure of route quality to guide the decision. The messaging protocol does have the capability for a device to issue a handset delivery receipt back to the carrier whenever a message is successfully delivered. However, those delivery receipts are not universally supported by all carriers or adopted in all regions. Worse yet, carriers can distribute fake receipts in some cases.

Cost, while a much more reliable data point, was not the primary measure Nexmo intended to use. Least-cost routing was already a standard practice in the industry for route selection, but Nexmo intended to put quality at the core of their offerings.

“So, then how do you measure the best route,” Nadalin asked before immediately answering his own question. “We go to our customers. We ask them to give us feedback.”

Some Nexmo customers were sending messages that prompted recipients to respond in ways that confirmed the messages had been delivered. Customers who used two-factor authentication (2FA) were prime examples. In a typical 2FA scenario, the Nexmo customer would send a new user a PIN code. The user would enter the sent code to activate the customer’s app or to log into their website. This action by the user closes the delivery loop on the sent message; the PIN code can be entered only if the message was delivered.

phone verification

Nexmo customers agreed to share their 2FA confirmation data, providing valuable quality insights about the routes through which those messages were delivered. “So we started to get this data point to apply to this inventory,” explained Nadalin. “That’s when Adaptive Routing started to come into the mix … let’s try to change the routing in a dynamic fashion depending on the feedback we are getting from our customers.”

“That’s how Adaptive Routing was created.”

Nadalin also pointed out that the gains Nexmo was able to make in deliverability, latency, and stability didn’t benefit only the customers who shared their data. The solution was applied to all of their customers’ messaging traffic, whether a social app company was using text to acquire a new user, or a financial services company was sending a one-time password to confirm a transaction, or a retailer was launching a massive marketing campaign via text.

What’s Next for Adaptive Routing?

Today, the Nexmo platform processes more than 60 million data points per month from customer feedback. The traditional carrier-provided delivery receipts are a productive input — delivering some 400 million data points per month — but the addition of delivery confirmations from actual end users provides Nexmo with a robust understanding of end-to-end performance. Ultimately, the combined data drives the Adaptive Routing optimization of SMS delivery.

Nadalin is not content to merely fine tune the patented solution though. He has bolder ideas for Adaptive Routing. “Where I’d like to go next is to combine the historical data [receipts and confirmations] with some predictive data to do some machine learning as well,” he said.

He envisions the Adaptive Routing algorithm processing the historical data and simultaneously receiving input from the predictive data as it selects routes. When the two data points suggest conflicting actions, Nadalin would like to alternate between the two and see which yields better results over time.

Learn more about Nexmo’s SMS solution to witness Adaptive Routing first hand.