Subex Juggles a Wide Variety of Business Assurance and Big Data Analytics Use Cases

In years past, Revenue Assurance (RA) used to be a fairly quiet professional niche with a high focus on stopping revenue leaks and isolating errors in ordering, provisioning, and billing.

That crucial role for RA is not going away, and yet operators are now calling on RA pros to serve up an expanding variety of business assurance programs to:

• Deliver greater inter-application and inter-department agility in the back office;
• Support subscriber scaling and the explosion in mobile and digital services;
• Boost RA, fraud, and security controls for emerging 5G and IoT; and,
• Supply analytics or machine learning/AI solutions to improve efficiency.

So it’s no surprise that RA’s heightened role has complicated life for the assurance software and services firms.  Supporting all the new cases has turned RA into a bit of a circus requiring software vendors to juggle their offerings and resources like never before.

One software player in the thick of this business assurance action is Bangalore-based Subex.  I caught up with Rohit Maheshwari, Head of Strategy & Products at Subex.

In my interview with him, Rohit provides a fine overview of the hot assurance and analytics solutions Subex is pursuing.

	Dan Baker, Editor, Black Swan: Rohit, I admire the flexibility of software firms like Subex who support telecom assurance and analytics.  As an analyst following RA, it’s often been tough to figure out where the parade is going.  But that’s nothing compared to the challenge of leading hundreds or thousands of employees to deliver products for this dynamic marketplace.

Rohit Maheshwari: Thanks, Dan.  Yes, it is quite a challenge especially as operators bring on new technologies and their profit margins get tightened.  To get our strategy right requires lots of planning, conversations with customers, and tinkering with new applications and platforms.

At Subex, we see opportunities in terms of three concentric circles (or horizons) that surround our business.

The nearest horizon is our traditional expertise in revenue assurance, fraud management, and partner settlement.  These solutions are the core of our business.  And we are quite bullish about this business and continue to invest in it.

On the second horizon are the solutions we have invested in over the past few years.  Security for the Internet of Things (IoT) is in that category.  These investments have started generating reasonable returns and we expect the returns to accelerate in the coming three years or so.

Finally on the third horizon are the longer range solutions that will leverage our pedigree and skills in AI/machine learning and data analytics.

	Scanning your website, I see a great number of use cases.  Can you give us a quick overview of some of the hot and promising ones?  I think it will give readers a good idea of where assurance and analytics is headed.

Yes, this a good way of illustrating trends.  Across the board, maybe the biggest overall trend is more expansion of the assurance practise into business assurance impacting all parts of the business and not just the revenue chain.  And for us, that means looking at new kinds of data and new use cases.

One example: not long ago, the revenue stream at large telcos came mostly from voice and SMS services.  However recently this has moved to device-sales — roughly 50% to 60% of telco revenue in some cases is now for handset sales and other devices.  And that’s a dramatic shift.

This means clients are increasingly asking for business assurance solutions in logistics, supply chain, and related areas from us — and less work in RA traffic reconciliation.

So let me walk you through a few of the leading use cases we’re seeing:

1.  Security for IoT

Let’s take connected cars, for example.  We increasingly see cars being embedded with e-SIMs and SIMs where there’s a risk of security breach.  In fact, several research studies show IoT security breaches are the number one barrier for adoption of Internet of Things (IoT).  So, the aim of Subex IoT is to build the world’s largest repository of threat signatures and provide security to a number of connected devices all around the world.

This solution takes Subex into a completely new domain — security.  And our solution will bridge across telecom, industrial automation, automobiles and other industries.  So, we’ve been pouring R&D into IoT security.  We are doing some of this work in collaboration with NTU, the Nanyang Technological University of Singapore.

2.  Optimizing the Use of Network Assets

Telcos are keen on spending money wisely in network expansion.  So, our Asset Life Cycle Management solution is about earning a telco a better return on their enormous network capital spend.  And in this area, we are working with a Tier 1 carrier in Australia and large telcos in North America.

For instance, if a card or multiplexer is available in the warehouse at point A, can it be re-purposed for use in point B instead of releasing a new purchase order?  Asset lifecycle management queries financial systems and the physical network to identify assets that can be re-purposed or moved to a better location to support, say, providing greater network capacity to a stadium or convention center.

3.  Big Data Analytics & AI Use Cases

Usage volumes continue to explode, especially at Tier 1 operators.  So, in the last few years, we invested in really scaling up our solutions, working on big data and Hadoop deployments to meet the demand.  Our largest deployment, for example, processes over 24 billion transactions each day, and is designed to continue scaling up to about 40 billion transactions per day.

As transaction volumes grow, doing analytics using traditional rules, thresholds, trending, forecasting, and these kinds of things gets much harder to manage.  This is why we’ve ramped up knowledge of machine learning, AI, and deep learning.

For us, it’s two distinct kinds of AI.  First, there’s the highly focused AI like SIM box detection where you are looking for very specific anomalies in data.  And let’s call the second a more general-purpose AI where we mine humungous data sets to discover outliers that lie on the edges of the distribution curve.  Here, it’s not just detecting the outliers, but also correlating one outlier with other signals in the data to create a business-incident-hit-list of problem areas.

4.  Massive Multiplayer Online Gaming Assurance

Another interesting case borrows from the gaming industry where massive multi-player online (MMO) gaming is a big deal.  Some of our telco customers now provide network-based play-along games and contests.

Think about the developing world where the mobile phone is often the only connected device and people are using the internet to watch T.V. and other content on an iPad or mobile device.  This is a massive trend here in India, thanks in part to the popularity of Bollywood content.  India is now home to some of the largest 4G LTE video and television streaming services in the world.

Say a consumer is watching a live sporting event on a telco streaming service.  Well, he’s encouraged to compete with other consumers in predicting outcomes in the sporting event.  Then, if the consumer predicts correctly, they score points or even land some cash prizes, free service, and so on.

Now this is a completely different kind of assurance.  You’re talking about a huge number of people interacting in real-time where the role of assurance is to ensure the customers playing the game are properly identified for awards, prizes and so forth.  This is not just about business assurance accuracy, but also it has to be done in near real-time.

5.  Assuring the Rapid Onboarding of Mobile Subscribers

Dan, let me walk you through a case study of ours involving a Tier 1 4G/LTE operator entering a saturated market.  The company has been in the market for 2.5 years, and then embarked on a campaign that had them scale in only 180 days to 100 million subscribers.

Now to enable hyper-growth in a highly saturated mobile market, the operator realized their number one priority was ensuring customers were getting the services they requested.  Specifically, they wanted to make sure when they onboarded customers, the services the customers subscribe to were all enabled in real time — and that breakages, if any, were rapidly identified in real time and remedial action taken.  We are very proud of the fact that this world’s largest greenfield startup’s initial customer experience was enabled by our Revenue Assurance solution.

Now this ability to identify breakages and respond so quickly led to a big competitive advantage.  At their peak, they were on-boarding nearly a quarter million subscribers every single day!  And the customer was assured of first time right services by us.

6.  AI-Powered Customer Credits & Complaint Analysis

One North American Tier-1 client is spending tens of millions of dollars every year in customer credits.  These are credits given by contact center agents when customers complain about a problem in service or plan not being advertised or understood properly.  So as customers complain, contact centers reps often pass on a credit of 10, 15, or 20 dollars to keep the customer happy.

Now the head of business assurance at the operator wanted to make sure credits were being applied by the contact center correctly.  And whenever contact center agents issue a credit, they would type a note in a memo of their application.

These memos were unstructured, free-flow text so it was not easy to “translate” and categorize them.

So this is where we engaged the client with our machine learning and natural language processing capability to properly categorize the memos.  The results were two-fold, as we: 1) shed light on the effectiveness of how credits were being applied, and 2) measured the perceived quality of experience for customers.

The operator actually found a rather high number of credits were being passed to business customers.  So this unexpected discovery enabled them to isolate the issue and take corrective action.

	It’s an impressive set of use cases.  And working with Tier 1 operators in these areas must require a fair bit of coordination as they develop internal solutions that complement or build upon frameworks you supply.

It’s true.  Tier 1 operators have started investing in and scaling up their own data science infrastructure.

Take fraud management as an example.  We continue to get value from the rules-based fraud control infrastructure we’ve grown over the years.  But we also want to create a machine learning or AI pipeline within the product so once data is consumed and stored in the system, we can extract features to build a machine learning model, then deploy these models into the fraud management system.

The trick then is to make our framework so open and consumable on the machine learning side that if the telco or the customer has its own strong data sciences capability, they can develop their own models and use cases, then deploy them straight into the FMS.

Now a big part of that is supporting enablers such as open source methods and computer languages such as R, Python, and H2O.

	Thanks, Rohit.  Subex has been very busy in the last few years.  And “busy” is a good place to be.  As the old proverb says: “If you really want something done, ask a busy person to do it.”