How to achieve a functional and effective BI service line for a Telecom Service Provider?
Based on the interest generated by my earlier blog on BI needs for a Telecom Operator http://www.infosysblogs.com/oracle/2009/04/an_insight_into_the_business_intelligence_needs_for_a_telecoms_operator.html#more
that I had written a few weeks back, I would want to take my view point a step forward. In my earlier blog I had mainly discussed about the possible stacks or spaces within a telecom operators landscape that needs a strong and well defined BI strategy.
In this blog I would want to discuss on some important architectural and functional tenets within which a Telecom players’ Business Intelligence roadmap should be architected and constructed, in order to derive maximum mileage.
Establish an Enterprise Data warehouse (EDW): It is very important to adopt an Enterprise Datawarehouse approach as opposed to creation of a set of federated BI structure, having dependant data marts. Such an architectural framework would lead to the creation of information silos, making information integration a huge challenge. Such a framework becomes a hindrance to establishing a 360 degree integrated view of the entire business like for example getting an integrated view of customer- account- revenue-inventory–network across a plethora of enterprise applications. So establishing an isolated data hub, a data mart and independent reporting areas is a big no no!
Adopt a standard EDW data model: This is something which I strongly believe is a “game changer”. The robustness, scalability and flexibility of an EDW is based on the adaptation of an industry standard data model. With more and more COTS based offering in the telecom space it is imperative that the EDW is also based on standards laid down by bodies such as TM Forum. A model based on TM Forum’s SID is slowly becoming the de facto data standard in the Telecom world. Whether EDW data models are purchased from mega-vendors like IBM, Oracle and Teradata or whether a home grown and customized data model is built, it is important that it is based on standards like SID. This approach will help you lay the foundations to establishing a high degree of interoperability and a relatively seamless 360 degree data integration between the various enterprise entities
Establish an enterprise ODS: With enterprise functions being spread across various applications and the demand for data movement between various apps being there, it is also important to ensure that a single version of the truth is maintained and used. For example the loyalty management application would need the revenue data from the billing system to calculate loyalty scores. More often than not it is the EDW which lands up acting as a feeder to downstream systems. Here I believe it is important to insulate the DW from such data demands and create a separate enterprise ODS hub to feed any such data subscriber. Although structurally a separate block but it makes all the sense to have this constructed and maintained within the purview of the Enterprise BI team.
Implement a consistent Data Quality Strategy: One of the key drivers behind ensuring a 360 degree enterprise view and seamless data integration is to establish the capability to match the primary identifiers across the various enterprise applications. For example: to identify the network elements usage by a large B2B customer and the corresponding revenue generated by such accounts would need an accurate data integration across the NOC, Order Management, Billing and CRM apps. The best place to do this and to get maximum benefit is the EDW and for that to happen well it would need a strict data quality level in the EDW. The key for being able to do this kind of integration accurately is being able to have cleansed, consistent, standardized and accurate data from the applications loaded into the EDW.
Get Real Time where needed: Typically the Campaign, Sales and some CRM apps would typically sit on top of an EDW with a dependant data mart created for these functions. It is important that the sales and campaign cycles are run with THE current data to have maximum effect and prevent ‘mis-targetings’. Wherever possible attempt should be made to build a Real Time replication mechanism and make some of your critical dimensions in the EDW refreshed on a real time basis. In case technology costs become prohibitive or a bottleneck, it would be advisable to make these integrations as real time as possible using CDC or real time replication methods offered by DBMS vendors like Oracle, Microsoft etc.
Look for and use standard packages wherever it fits the needs: Telecom players have certain very specific analytics needs like for example Revenue Assurance, Network Inventory optimization, Churn analysis( which is discussed below as well).Most of these areas are addressed by use of packages which mine into an EDW. Here my take would be to adopt the offerings from industry leaders to take advantage of well established process and data mappings and well as time tested models to help address the business issues. And these packages plug in well on top of a standardized EDW.
What lies ahead?: While the mystery of “What lies beneath?” is more or less handled by having an robust analytics layer on top of your EDW, the issue of “What lies ahead?” is handled with establishing a team for performing a data mining and predictive analytics. Data mining and predictive analytics is what would help to take your BI initiative to the next level and is very critical for a competitive telco. This is especially true in the business of Telecoms where predicting churn, predicting a competititive penetration, predicting new product success ratios are key. In this regard it is important to make the right choices with regards to procuring and effective mining tool, establishing and fine tuning models and bringing in the right team from the various enterprise functions to work on this niche function.
In an industry which has to adopt technology at a break-neck speed to be ahead in the game and stay afloat, some principles of establishing an effective BI function are often compromised and lost sight of. However ailing to certain fundamentals which I have tried to bring forward, the overall strategy would go a long way in ensuring a robust, effective and functional BI service line.
Adopt a standard EDW data model: This is something which I strongly believe is a “game changer”. The robustness, scalability and flexibility of an EDW is based on the adaptation of an industry standard data model. With more and more COTS based offering in the telecom space it is imperative that the EDW is also based on standards laid down by bodies such as TM Forum. A model based on TM Forum’s SID is slowly becoming the de facto data standard in the Telecom world. Whether EDW data models are purchased from mega-vendors like IBM, Oracle and Teradata or whether a home grown and customized data model is built, it is important that it is based on standards like SID. This approach will help you lay the foundations to establishing a high degree of interoperability and a relatively seamless 360 degree data integration between the various enterprise entities
Establish an enterprise ODS: With enterprise functions being spread across various applications and the demand for data movement between various apps being there, it is also important to ensure that a single version of the truth is maintained and used. For example the loyalty management application would need the revenue data from the billing system to calculate loyalty scores. More often than not it is the EDW which lands up acting as a feeder to downstream systems. Here I believe it is important to insulate the DW from such data demands and create a separate enterprise ODS hub to feed any such data subscriber. Although structurally a separate block but it makes all the sense to have this constructed and maintained within the purview of the Enterprise BI team.
Implement a consistent Data Quality Strategy: One of the key drivers behind ensuring a 360 degree enterprise view and seamless data integration is to establish the capability to match the primary identifiers across the various enterprise applications. For example: to identify the network elements usage by a large B2B customer and the corresponding revenue generated by such accounts would need an accurate data integration across the NOC, Order Management, Billing and CRM apps. The best place to do this and to get maximum benefit is the EDW and for that to happen well it would need a strict data quality level in the EDW. The key for being able to do this kind of integration accurately is being able to have cleansed, consistent, standardized and accurate data from the applications loaded into the EDW.
Get Real Time where needed: Typically the Campaign, Sales and some CRM apps would typically sit on top of an EDW with a dependant data mart created for these functions. It is important that the sales and campaign cycles are run with THE current data to have maximum effect and prevent ‘mis-targetings’. Wherever possible attempt should be made to build a Real Time replication mechanism and make some of your critical dimensions in the EDW refreshed on a real time basis. In case technology costs become prohibitive or a bottleneck, it would be advisable to make these integrations as real time as possible using CDC or real time replication methods offered by DBMS vendors like Oracle, Microsoft etc.
Look for and use standard packages wherever it fits the needs: Telecom players have certain very specific analytics needs like for example Revenue Assurance, Network Inventory optimization, Churn analysis( which is discussed below as well).Most of these areas are addressed by use of packages which mine into an EDW. Here my take would be to adopt the offerings from industry leaders to take advantage of well established process and data mappings and well as time tested models to help address the business issues. And these packages plug in well on top of a standardized EDW.
What lies ahead?: While the mystery of “What lies beneath?” is more or less handled by having an robust analytics layer on top of your EDW, the issue of “What lies ahead?” is handled with establishing a team for performing a data mining and predictive analytics. Data mining and predictive analytics is what would help to take your BI initiative to the next level and is very critical for a competitive telco. This is especially true in the business of Telecoms where predicting churn, predicting a competititive penetration, predicting new product success ratios are key. In this regard it is important to make the right choices with regards to procuring and effective mining tool, establishing and fine tuning models and bringing in the right team from the various enterprise functions to work on this niche function.
In an industry which has to adopt technology at a break-neck speed to be ahead in the game and stay afloat, some principles of establishing an effective BI function are often compromised and lost sight of. However ailing to certain fundamentals which I have tried to bring forward, the overall strategy would go a long way in ensuring a robust, effective and functional BI service line.
