|© 2016 Black Swan Telecom Journal||•||protecting and growing a robust communications business||• a service of|
|Email a colleague|
In case you haven‘t noticed, a bunch of software startups have exploded onto the telecom analytics scene in recent years. So the question is: will this business continue to grow or will it fizzle as soon as IBM stops its “big data” and “smarter planet” ad campaign?
My sense is that the rise of telecom analytics is very real. In fact, the availability of low cost computing power has opened the flood gate to dozens of analytic apps that deliver real value for telcos. Analytics gurus talk about three V’s that drive analytics opportunities: the Velocity of the business, the Volume of data you have to analyze, and the Variety of data. Well, telecom has all those qualities in abundance.
And by virtue of its thirty years of experience in the market, you can bet that SAS Institute will continue to be a big beneficiary of this trend.
SAS Institute should be no stranger to those following high tech. The North Carolina-based private company with a global reach is today’s largest player in telecom analytics with communications industry revenue accounting for 7% of its $2,275 million total (2011).
Rated by Fortune magazine as one of most employee-friendly companies in the U.S., the company is a solid business too: despite the recession, SAS grew its revenue a healthy 11% from 2010 to 2011. Steady at the helm, SAS has stayed the analytics course and grown the business organically while its competitors spent billions in acquisitions to get into the same race.
Another thing I like: SAS reaches out to editors and potential customers in a down-to-earth way. Check out their Customer Success webpage with its hundreds of cases studies-a full 29 of which are cases in the comms industry. Each case is professionally written with a mind to educate the reader with food for intelligent thought.
So we’re pleased to welcome long term industry colleague Ken King, Director of Telecommunications at SAS, who gives us a valuable historical and future-looking perspective on telecom analytics.
|Dan Baker: Ken, looking back to the day when SAS was one of the very few companies you could classify as a true “analytics firm”, I’m wondering how the arrival of “big data” or low cost computing power has changed the way analytics is done.|
Ken King: Dan, I think the big difference is around how fast you can do the analysis. When you can enable these things to run in memory, your models can be updated in minutes instead of hours. So instead of updating those models quarterly, you can update them perhaps daily or even multiple times a day.
As we look at the models we’ve built in the past, they run reasonably well, but they required a filtering process that causes you to move a lot of data in and out. For example, in the past we would filter and analyze the top 1% of something because that’s all the customer could afford to process. Today we can avoid much of that intermediate processing and put everything in memory and process it very fast.
That means we are increasingly looking at all the data rather than a sample. So that gives you a big lift because looking at all of the data is better than the sample. But if there’s no or relatively little extra cost to look at all the data versus just a sampling, then using all the data is better because you can look at all the outliers.
|So what’s new at SAS Institute in telecom? What part of the market are you seeing some potential growth in?|
Over two dozen of our telecom customers are using the product of ours we call Customer Link Analytics. What this does is reveal which customers have the largest social circles. It recognizes those people who are better connected and are more influential -- or if not influential, then certainly people provide an early indicator of trends.
Traditionally the techniques for reducing churn have been based on algorithms that only looked at that customer’s individual records: it didn‘t look at the social network that a customer was involved in. Customer Link Analytics adds an entirely new dimension to the tools used by marketers for retention and launching new products and services.
Doing this analysis for millions of customers is enormously compute-intensive, so not many companies could afford that capability up to now. I mean, imagine a big spreadsheet where you had every phone number on one side and every phone number again on the other axis — and as start connecting one phone number to all the others, it plots out as a huge, sparsely populated matrix. Later this year we will release a version that is enabled for in-memory analytics. So the processing time will drop from many hours to just a few minutes.
|I believe this technique is called “social networking” and it is totally different, of course, from the social network we talk about in the context of Facebook or Twitter.|
That’s right, and it does cause confusion, though I’m not sure exactly what we can do to eliminate that confusion.
Actually, social network analysis goes back a long way in telecom. For instance, least cost routing was done using this technique and the idea there was to see how many calls were passing through a particular intermediate node in the network. And in this way you could figure out where spending your next dollar in network would do the most good.
So those were all sort of ways of social networking and then all of a sudden Facebook and LinkedIn came along and they sort of hijacked the terminology. What’s different about the social network analysis I’m talking about it’s done using your own call detail records: it’s data that only you have access to and it is a 100% objective data. Everything that you are looking at has something to do with your customer using your products. There certainly is an analytic opportunity to look at comments posted on Twitter and Facebook, but that’s the exact opposite. With Facebook comments, the data that is not just available to you; it’s available to anybody. And that data is also very subjective for the user and you have to filter out the 99% of information that has absolutely nothing to do with your products and services.
|Can you give a quick example how you would utilize Customer Link Analytics?|
Sure, Dan. Let’s say we’ve launched a new Android or iPhone app and we want to observe how quickly it propagates through the network. Well, we start seeing that app being used by the influential users and pretty soon the next group of people downloading the app are closely connected with them. That’s a very good indication that the app is successful and will probably do well, so based on that info, maybe we should do some marketing campaigns to promote the app.
By the same token, if you notice that your app isn‘t propagating, well, that’s a signal that the app is not very good because the key influencers are not telling their friends about it.
|Ken, as an analyst, I’m always trying to sort out the markets I study. So what about analytics? How do you carve out the pie? Do you internally track a “customer churn prevention” segment vs. a network optimization one?|
Churn is a different animal I think. I mean, almost everything an operator does is in some way related to reducing churn. Churn management is much more than creating a model that says who is likely to churn. The important part is going back and addressing the factors that are causing the churn. And this opinion is reflected in our own surveys when we ask customers why they bought our software. So while churn management is certainly a permanent issue, its causes are many.
As we see the telecom analytics market, there are three major segments: customer intelligence, information management, and network analytics. These are the top sectors, but then there’s what we call service optimization which includes a bunch of different things such as risk management, which includes revenue assurance, fraud, and margin assurance.
Another area is “analytics model management”. A company may have thousands of different analytical models put together, so how do I manage those? And each is in a different maturity phase across its lifecycle. Some models are in test mode. Others are in production. Still others need to be sent out to pasture because they are no longer useful.
So model management entails monitoring the effectiveness of models. When someone comes along and says, “I’ve got a great idea for a model”, well, you can look back and see: have we done anything like that before; do we actually already have it; and what results came out of that model? In short, as soon as you begin collecting a large number of analytical models, it’s easy to get mixed up unless you’ve got a good process for managing it.
|Coming into the analytics business in the very early years, you’ve prospered without tying yourself to any particular hardware religion.|
Yes, we have always been very much hardware and database independent. And that kind of brings you back to the thing that has made SAS so successful — the ability to get data from anywhere. Talk to some of the people who have been here a long time and they will say: “We have never seen a data source we couldn‘t access.” So, you are right, we are independent.
And the way the market is evolving, you’ve got big providers like Oracle and IBM who are betting they can be a one-stop shop — buy hardware and the software from the same vendor. SAP is another company doing lots of acquisitions and acquiring that entire end-to-end thing, but our strategy at SAS is to stick with what we do well, which is the analytics — and supporting all the platforms that we believe are important.
|How does SAS typically engage with telecom industry clients? Are you going in and doing some upfront consulting and leaving some analytics software behind?|
We have many, many different models of business that we do, but perhaps our most common scenario is to engage with a customer about its hot-button issues and how we could help them solve those issues.
Obviously over the last couple of years, the big trend in wireless has been the movement away from feature phones that make calls and send text — and towards the smartphone. So the rise of the smartphone means that a lot of analytic models need to be recalibrated because people are now using apps, they are a lot more data dependent and so forth.
So, we come in and try to understand: what are the factors that are changing those models? We often start off by doing a proof of concept, showing how it would work and then integrating it into existing systems. So, the customer is then licensed to use the software for that particular process.
The customer gets an end-to-end analytical application that provides them answers to a business problem and there are different models of how we do this. In most cases the client then takes over and their people run and maintain that system.
Often the customer has its own experts who‘ve been SAS users for a long time and they know exactly what they need to do and just use a software license to do it. Other times they need various amounts of assistance from us.
In some cases , we’re like the home improvement store where somebody shows up at the cash registrar with a bunch of boards, saws, nails and hammers and they walk off and go build something. We often have no idea what they have built until they come back and might say, “Hey, I have built this neat thing and I would like to do a presentation about it at the SAS Global Forum.”
So when those models happen and a customer shares something he’s done, we and our customers learn from that. But sometimes they come in and say, “Gee, I want to build a dog house but I don‘t really know how to get started”, and so we work with them to build a complete business solution.
The majority of our revenue comes from the follow-on software license. To keep that revenue stream flowing we have to continue delivering value to our customers. So we need to understand their business challenges and work with them on solutions.
Copyright 2013 Black Swan Telecom Journal