With the planned sunset of the Developerworks blogs, all of my past content will soon be disappearing from the Internet.
I'm saddened to read this, so I'll be creating a new blog and will copy some of my past blog posts to that platform. As I'm no longer an IBMer, that's unlikely to be the centralised IBM platform that is driving the elimination of this platform. Once I have a new blog platform with some of my old content posted there for posterity sake, I'll update this post too...
(OK, it's not strictly Telco related, but check the footnote to see my personal connection with the J9 VM in particular)
WebSphere Liberty is the high performance Java Enterprise Edition Server that's ultra light-weight - it includes and OSGi container and uses the IBM J9 VM at it's core - which IBM has also donated to the Open source community (via Eclipse - see https://projects.eclipse.org/projects/technology.openj9) .1
1. A little history lesson on the IBM J9 VM - it was originally developed by IBM's (now defunct) Pervasive Computing division for IBM's J2ME lightweight VM. It was then ported to the J2SE and J2EE platforms. When it was developed, I was in a Tech Sales role for the Pervasive Computing division, so I have a soft spot for the J9 VM.
An ex-colleague of mine (Violet Le - now the Marketing Director at Imageware) asked me about the drivers for Analytics in Telcos. I'll admit that it's a subject that I haven't really given a lot of thought to - all the projects that I've worked on in the past that have included Analytics have had a larger business case that I was trying to solve... Marketing, Future Planning, Sales etc I've never worked on an Analytics project for the sake of analytics, nor have I designed a solution that was just (or mainly) analytics.
There is a definite value in analytics in providing an insight into how the business is running - to enable business to plan for the future and to manage how they run in the present. Both Strategic and Tactical cases for analytics would seem to me to be of value to any business. An analytics system that delivers insight into the business (customer behaviour, sales effectiveness, capacity usage and predictions etc) is great, but at the end of the day, a Telco needs to do something about that information/insight to actually deliver business benefits.
As I'm no analytics specialist, I wont' try to describe how to define or build those systems. What I will try to do is to describe the bits around the analytics systems that make use of that insight to deliver real value for the CSP.
What are the business cases that I've seen?
Sales & Marketing
Driving promotions to to positively affect subscriber retention or acquisition... I did a project with Globe Telecom in the Philippines which was primarily aimed at driving SMS based outbound marketing promotions that are targeting based on subscriber behaviour. An example might be if a subscriber had a pre-paid balance less than (say) 5 pesos, and the subscriber topped up more than 20 pesos and less than 50 pesos, then send a promo encouraging the subscriber to top up by more than 100 pesos... all the interaction is via SMS (via a ParlayX SMS API)
Back in 2013, I did an Ignite presentation at the IBM Impact Conference in Las Vegas - Here is the presentation (Smarter Marketing for Telecom - Impact 2013) The session was videoed, however, the video recording is no longer on Youtube. Happily, I've found the video is still available - just not easy to find. Here it is for your enjoyment!
Social networking analysis to determining who should be targeted. IBM's Research group was pushing for years a Social Networking Analysis capability that looked at Social Networking connection to determine which subscribers are followers, which are community leaders and influencers and based on that assessment.
Ensuring utilisation of the network is optimised for the load requirements. I worked with a telco in Hong Kong that wanted to dynamically adjust the quality of service level to be delivered to a specific user based on their location (in real time) and a historical analysis of the traffic on the network. For example, if a subscriber was entering the MTR (subway) station and the analytics showed that particular station typically got very high numbers of subscribers all watching youtube clips at that time of day on that day of the week, then lower the QoS setting for that subscriber UNLESS they were a premium or post-paid customer in which case, keep the QoS settings the same. The rating as a premium subscriber could be derived from their past behaviour and spend - from a traditional analytics engine.
Long term planning on network (SDN/NFV will allow Networks to be more agile which will reduce the need for traditional offline analytics to drive network planning and make the real time view more relevant as networks adapt to real time loads dynamically ... as traffic increases in particular sections of the network, real time analytics and predictions will drive the SDN to scale up that part of the network on demand. This is where new next gen AI's may be useful in predicting where the load will be int he network and then using SDN to increase capacity BEFORE the load is detected... read Watson from IBM and similar....
A few years ago, a number of ex colleagues (from IBM) formed a company on the back of real time marketing use case for Telcos and since then, they've gone ahead in leaps and bounds. (Check them out if you're interested, the company name is Knowesis) <edit> Unfortunately, I can't link to them without setting off the spam protections on mydeveloperworks, but I'm sure you can figure it out... www.knowesis.com will do it!</edit>
Do you have significant use cases for analytics in a CSP? I'm sure they are and I'm not claiming this is an exhaustive list - merely the cases that I've seen multiple times in my time as a solution architect focused on the telecommunications industry.
I wouldn't normally just post a link to someone else's work here, but in this case Frank Wong - a colleague of mine at my new company (DGIT Systems) has done some terrific work in helping to eliminate the miss-match between the data model used by the TMF's REST based APIs and the TMF's Information Model (SID). I know this was an issue that IBM were also looking to resolve. In the effort to encourage the use of a simple REST interface, the data model used in the TMF's APIs has been greatly simplified from the comprehensive (some might say complex) data model that is the TMF's Information Model (SID). This meant that a CSP who is using the SID internally to connect internal systems needed to map to the simplified API data model to expose those APIs externally - there was no easy one-to-one mapping for that mapping which meant that the one could not simply create a API for an existing business service (eTOM or otherwise) - a lot more custom data modelling work would be required.