Informational Analytic Systems for Telecoms. Who Owns Information, Owns Control over Situation.

Interview with Alexander V. Ivakhin, CBOSS Group Vice President

Billing and CTI magazine #1/2003

At III International forum “Billing. IT Telecom”, held in December 2002 in Moscow, CBOSS Group received Product of The Year award in the sphere of analytical systems for telecoms. Read the interview with Mr. Ivakhin for an insight into the infrastructure of information-analytical systems deployed at a modern telecommunications operator.

Alexander Vasiljevich, could you please highlight major constituents of IT infrastructure, typical to a modern telecom enterprise and their strategic value for business?
Any telecom enterprise, in the course of business, generates huge data flows. Some part is used in current operations, for example in billing, network management and CRM transactions. Other data flows are not utilized and get lost. However, all information could be used to advantage for analyzing of accumulated experience, current state of the system, assessing consequences of  made decisions, thus making business plans and forecasts more proactive and successful.
Modern IT enables handling any sort of information, registered by the system that could be used to increase overall enterprise performance efficiency. For better understanding of how it works, it might be helpful to consider a telecommunications company IT infrastructure.
Let us consider IT as a multi-storey building, each floor representing an underlying technology. The basement belongs to OLTP (On-Line Transaction Processing) that includes Operational Resource Management systems [ORMS], such as, in case of a cellular operator, MSC, SMSC, VMS, etc.
OSS (Operation Support Systems) build on the next floor, on the basis of ORMS. OSS address the needs of mediating between ORMS and business users. Operation Support Systems support business processes and integrate them into a single information medium.
Next floor of the infrastructure realizes TMN systems that provide operative control over the lower levels, that is ORMS and OSS. TMN systems implement OLTP technology and monitor performance of the systems of the lower levels, tracking their status and alerting on all sorts of events happening in the system. Ultimate goal of these systems is provision of error-free, fault-tolerant operation of solution in general.
The first three “floors” serve the basis for analytical decision support systems (DSS) that implement OLAP technology. Let us consider CBOSSdss, an example of such solution. The main purpose of technologies, implemented on this layer is collecting and accumulating of data, received from the lower levels of ORMS, OSS and TMN. After processing and aggregation, the data is presented to business a user in form of analytical information that provides an insight into various aspects of enterprise performance, for example, customer relations, settlements, provided services, etc.
Provided example is typical for implementation of an infrastructure pattern by a telecommunications company. Moreover, it complies to recommendations of the TMN model. Most companies implement prescribed levels, or «floors», however, hi-tech interaction of the realized systems becomes a challenge.

What solutions from CBOSS product line implement IT “floors” that you have mentioned?
Our product line comprises over 30 systems, and I believe it would be inappropriate, in the course of the interview to name them, to say nothing about classification and functionality overview. Suffice it to say that we provide integrated solutions for all levels that bring automation of information flow utilization, making data handling comfortable and ergonomic.
It should be noted, that Management Information System, represented our product line by CBOSSmis is on the top of the four levels of the “IT pie”, that I have already mentioned. It is noteworthy that at the current stage of market development, MIS systems have become a crucial factor in achieving and retaining competitive advantage, which is vital for growing businesses. Control and manageability of a company becomes a true challenge that can be solved enterprise-wide only with the help of information management systems. 

What is the place and role of a billing system in a telecom operator’s IT infrastructure? A billing system belongs to an OSS level, does not it?
Exactly. The core mission of a billing system is to receive reliable data from ORMS about actual events, for example about calls, to rate these calls and present a bill, generated on the basis of this information, to a subscriber. In a broader sense, billing includes adjacent business processes: when a bill is presented, it is necessary to make sure that payment is duly performed. Moreover, it is important to arrange facilities for receipt of payments, support service, etc. 
Naturally, that our core product, CBOSS, Convergent Operation Business Support System, includes besides a billing system itself, a number of OSS subsystems that bring automation to major business processes of an operator, such as conclusion of contracts, subscriber activation, customer care, financial control, etc.

Could you differentiate OSS and DSS levels from the process stream point? Well, making a decision is also a business process, is not it?
The principal difference of systems on OSS and DSS levels is different structures of automated business processes.
OSS level business processes are strictly determined and consist of a sequence of operations, that require consecutive execution: output of the previous operation is input data for the next operation, etc, until the final result is achieved.
DSS level business processes are different from the very start. They always include creativity elements of a DSS user, thus they cannot be patterned. For example, it is practically impossible to work out an unchangeable business process that would define the most profitable for an operator tariff. Naturally, in the course of user interaction with the system a business process will be worked out. However, the business process will be formed as a result of research and decision making and it will be possible to reproduce it for other tasks only in part.

Could you elaborate on support, delivered by an information system? Is it a multi-aspect reporting facility or a set of recommendations to be followed? What is the role of human factor in the process of analysis and decision making?
The dialog with information system, obviously, lacks formalization. I would call it research work. For example, any operational process has a strict sequence of actions. It has a starting point and an endpoint. A research process, on the contrary is better described as a graph, and you traverse edges, assisted by a decision support system, with an option to follow various paths. To put it differently, there is a starting point and a graph, representing a tree of solutions. The art of an analyst consists in choosing the correct path in traversing graph edges, and come to the right solution.

There is a an opinion, presenting an analytical system as a know-everything button, that gives the solution to your problem, once pressed…
This is a very misleading point of view. An illusion of a system that would provide ready decisions, like “Introduce a new tariff with the parameters given below…”, “Extend coverage in area X”, or “Launch an advertising campaign for service Z”, etc, will remain a never-to-happen dream.
Consider mathematical models. The system would require a lot of them. Suppose models that describe a situation on a telecom market, current macro-economic situation in this country, patters of subscriber behavior towards tariffs, new services, innovations, competitors on telecom market, etc. Complexity of the task can be compared to weather forecasting challenge that still remains unsolved. Even if such a model existed, it would require unthinkable computational capabilities and an academician-level intellectual potential from personnel for model processing.
There are a number of challenges in building AI based decision support systems. One of them is inability to make the system learn on the real subscriber base. Moreover, working with AI systems requires more analytic skills from personnel than with classic analytical solutions. However we cannot deny application of AI in some elements of analytical systems. I think that some day AI systems will become mainstream. 

Is it safe to say that an analytical system is a tool that produces recommendations rather than instant decisions?
Exactly. There is no ready-to-use decision at the system's output. However, the system provides a user with capabilities to process information from the lower levels of the «IT pie», while going a winding road of research, and retrieve information that would help make a decision. In other words, an analyst is responsible for the process of acquiring information, required for making a decision, but it is a duly authorized manager who will make it.  
An analyst has every tool needed for retrieving, filtering, and presenting information either in tabular or graphic form. A plethora of methods and selection criteria exist, and a user can choose those that match current analytic tasks best.
Consider our product, Decision Support System CBOSSdss. The solution addresses enterprise-scale provision of information and information-processing tools for managers and analysts. The system allows to carry out comprehensive, multi-aspect analysis of enterprise performance subject domain. For example, settlements with subscribers, dynamics of subscriber base, subscriber traffic, service volumes per distribution channel, subscriber groups.
CBOSSdss implements a modular architecture and enables us to create an array of datamarts, individually assembled for analytic demands of an operator. Upon customer request more datamarts are added smoothly. To date, over 40 datamarts are available, grouped in modules, each module targeting a specific aspect of an enterprise performance. For example, a financial module addresses the need of enterprise financial analysis: realization, payments, subscriber debt, a subscriber module targets changes of subscriber base, its geographic and demographic profiles, etc.

What data model the system implements?
Information for CBOSSdss end-users, as I have mentioned already, is stored in datamarts, consisting of structured set of multi-dimensional measures. A measure is included into a datamart according to its functional purpose. All measures, concurrently analyzed by a user with a client application, constitute a dynamically changing “hypercube”. A hypercube dimension is an attribute from subject domain, covered by a datamart.
CBOSSdss user handles a hypercube with slices. A slice is a subset of a multi-dimensional array corresponding to a single value for one or more members of the dimensions not in the subset. From an end user perspective, the slice is a two-dimensional page selected from the cube. Different slices allow end- users to view data from different points. 

What are the key enablers of a successful analytical system implementation?
To begin with, it is the attitude towards implementation project inside the company. In other words, the attitude shows how mature an operator is for implementation of analytic technologies. Before an implementation starts, we try to create a demand for such sort of systems. That is we show what kind of tasks such system would handle and what are the benefits of decision support system automation. 
Next enabler of successful implementation is availability of qualified personnel. An enterprise should have specialists, capable to utilize all functionalities the system provides. Besides analytical skills, future end-users should have a clear understanding of enterprise business. The satiation is worse, however, if a company does not have specialists with sufficient research skills or core tasks of analysts used to boil down to report viewing. In this case, inability to utilize every functionality is likely to result in frustration in analytical tools.
Consider a favorable scenario, when a company has promising analysts. A dedicated specialist of billing or marketing department, responsible for reporting tasks will do. Then, there is definitely more chances for a successful implementation.
Addressing the need of businesses for highly-qualified personnel, we offer special education courses.  Besides, in the framework of consulting services, we share with our customers the techniques we have, for example how to measure efficiency of a service, or how to make a new tariff, etc.
Finally, the third enabler is attitude of customer’s top management to the project. At the early stage of cellular telecommunication market in Russia, the dominant guideline for an operator was availability of a license. Subscriber growth came naturally with a mere provision of core services. The situation has changed, however. Now it is about available services, service packages, tariff policy, marketing strategy. Thus, it is about the tools that will handle such tasks. That is why we see a solid interest for analytical systems. If top management of an operator shares this point of view, it supports the project and predefines its success.

What about analytical systems and ROI? Do we have effective and efficient techniques to assess ROI in decision support systems?
This is a challenge that still has to be met. I would point out two factors in ROI assessment. The first one is about impact on bottom line and cutting costs. The second one is about synergetic effect of system implementation and ownership.
Bottom line effect is easy. Firstly, utilization of DSS decreases workload on the main database server. This makes performance requirements to datacenters less demanding. Indeed, end-users requests are processed on a dedicated data warehouse rather than OSS server. In its turn, DWH server loads information from OSS server when user activity is minimal, if any.     Bottom line effect can be really impressive. One CPU board for a hi-end server translates to real savings.
Then, it is possible to assess saving of work time for analysts and bound IT personnel. We possess such assessments and share them with our customers in the process of forming a demand.
Finally about synergetic effect. With analytic system at hand, having a tool for fast access to research data, an analyst is able to create a tariff that would be a gold mine for an operator.  Synergetic effect is really hard to materialize. Obviously, without an analytical system, working out a killer-app tariff would be impossible. This is the simplest example. Other examples come in swarms: tracking down an error or flaw, or inefficiency in a service. Metaphorically speaking, an analytical system is a monitor that highlights issues of business performance at their early stage, helps to solve them timely and proactively and thus avoid losses.
Although it is impossible to assess DSS implementation synergy in figures beforehand, we are confident, that its effect is overkilling, far exceeding the cost savings share.

My last question is about CBOSS success in implementing high-level technologies of telecom IT infrastructure. What is your vision of the success?
We declared and proved, repeatedly, that to date, high-level technologies become available for telecom operators only as part of an integrated solution. That is why our integration of CBOSSdss into a complex solution is a qualitative advance towards distribution of analytical technologies for telecommunications.
With a well-articulated demand for analytical systems, the crucial enabler of penetration becomes price. Analytical systems become an affordable purchase because they are shipped as part of solutions, integrated both horizontally and vertically.
Our company performed 12 installations of such solutions over the last two years. Such results would be hardly attainable with a stand-alone non-specialized analytic system that always requires individual adaptation.
To crown it all, we see that successful implementation experience, the ultimate indicator, proves the efficiency of the industrial, complex approach we chose for erecting an IT building.