By Heena Jhingan
With about 10% share of the Indian telecom pie, Aircel has come a long way from being a regional player in Tamil Nadu and now offers a slew of voice, data and enterprise solutions. The company, co-owned by Malaysia’s Maxis Communications, has emerged as one of the fast growing telcos in the country.
However, despite the pace of growth, the cellular operator till some time back lacked a single window view of its customers.
Such systems make the processes time consuming and it also becomes difficult to establish relationship between various parameters. The operator thus, needed an enterprise data warehouse (EDW) solution atop which they could integrate the business intelligence (BI) tools to make analytical inferences.
The warehouse strategy
The cellular operator evaluated several major EDW providers but having rated them on various technical, strategic and commercial factors, Aircel decided to close on Teradata as it offered a combination of these factors to match the company’s business needs.
Along with its IT partners Wipro, Aircel’s IT team set out on implementing the solution. Sewell informs that they decided to follow a phased deployment approach, the first phase commenced in August 2011 and went live in February 2012, followed by Phase 2 in October 2012 and finally they closed the third Phase in March this year.
Aircel Enterprise Data Warehouse was set up on Teradata Appliances. This, Sewell says was done carefully, keeping the processing and data warehouse layers separate so that ETL (extract, transform, load) processes do not affect the end user query performance. “Data gets processed on Teradata 5x/6x series boxes, which is then moved onto Teradata 2x boxes through data mover utility,” he explains.
Dinesh Jain, Country Manager, Teradata India believes since the telco operations are very complex, it is better to have different stacks for processes. “Aircel’s total analytical ecosystem amounts to about 250TB, in such scenarios it is important to keep it simple,” he stresses.
Once this was done, Sewell adds, “As part of Enterprise BI, other technologies were also used. We deployed SAP Business Objects, BO Explorer, BI on Mobile, a Hadoop based platform for faster processing of profile attribute and recommendation/campaign management tools.”
The implementation brought together various lines of business and over 20 varied data streams into a central data hub. Now with the integrated data view provided through the EDW/BI solution, the cellular service provider is able to get a comprehensive view of the customer life cycle, which includes tasks such as customer identification, acquisition, customer relationship management, retention and value enhancement.
According to Sewell, the implementation has eased the processes for not just one, but various departments. For instance, the device department is now able to classify customers based on device/handset type. With BI, KPIs such as device-wise revenue, data around ARPU, rotational churn and age on network are available to the business.
“Leveraging on the customer profiling framework, we have recently started implementation of predictive or statistical modeling, including churn modeling, cross sell/up-sell and revenue forecasting,” he explains.
Aircel’s sales and marketing team now closely maps KPIs related to network cell site as this allows them to monitor performance of each cell site and plan campaigns to extract maximum benefits from low utilized network sites. This has also helped in promoting geography based product constructs.
Since sales executives need to access data on the go, Aircel needed to invest on BI on mobile. “So we innovated with releasing the BI KPIs over SMS — this helps in pushing information to the sales executives on the field as and when needed,” he says.
The best part of the implementation is the availability of an integrated, centralized and online business view through BI, which eliminates the need to keep offline data repositories. This results in significant capex savings, covering risk of data security and substantial savings on off-role resources that were earlier required for manual data processing.
Challenges
Bringing about 20 data streams together is definitely not an easy thing. As Jain points out, data modeling plays a critical role in making the analytical tools effective.
Even though a group of customers might be using similar products and services, they might have different usage patterns based on their segments, geographies and devices used. Getting such insights can be an arduous task for telecom operators. Post excessive discussions and workshops, Aircel modeled its profiling framework.
Sewell says that a bigger matter of concern for them was the fact that their EDW was evolved parallel to the development of enterprise (source) applications. Any change in the flow or architecture of the source application had a direct impact on EDW project timelines. “We could handle the change because we managed to maintain synergies within teams through effective communication processes,” he reasons.
He adds that since data from many streams used to travel to the EDW/BI via multiple hops, ensuring data completeness and manual reconciliation was a mammoth task. Implementation of auto reconciliation at each data hops and alert triggering mechanism at various stages for data loading/processing was an innovative approach that has resulted in optimal data quality, user satisfaction and ease of manageability.
This being Aircel’s first EDW implementation was a reason that till now, the business users viewed the business intelligence and EDW key performance indicators (KPIs) from a mere operational perspective and this had to be changed. They involved a business analyst, functional SPOCs (single point of contact) and the EDW project life cycle team to overcome this challenge.
Fine tuning ahead
With completion of the third phase, Sewell informs that all major milestones have been achieved and they are now under an optimization phase. Following the initial implementation, Aircel undertook a couple of critical initiatives, including integrating the system with advanced analytical tools and implementing BI on mobile.
However, the operator believes there is still a lot more to build. Future plans call for building capabilities of processing unstructured data in the BI platform and enhancing social media analytical capabilities. Also in the offing is the integration of the platform with operational systems to close the loop with analytical outcomes, so that operational systems can save CPU cycles (the time in which a single instruction is fetched from memory, decoded and executed). The idea is to add business value at an operational level.