Traditionally, analytics have been applied in the telecom industry for Marketing purposes where Customer Segmentation was done to decide on what promotions to offer to subscribers and what they are likely to use. Further, customer loyalty and customer retention programs were also deployed using analytics. Over a period of time, analytics started impacting the telecom sector in more direct ways through Network Optimization solutions that allowed Operators to monitor the resources to help them plan the capacity upgrades.
Customer Care and Campaign Management like analytics solutions came from a backdrop of traditional marketing domain. While marketing originally was inserting ads in newspapers or distributing pamphlets, it has evolved into digital marketing with a clearly identified target audience. Initially, the “hit ratio” or “returns per dollar spent in advertising” was very low; but increased as the accuracy of predicting the target audience (thus targeted advertising / marketing). This yielded better results. This was acceptable too as the hit ratios bettered from typical <5% to 30–40%. This was a great increase in RoI. To put it directly, even low or medium accuracy of the outcome was acceptable as it was giving a big leap in the returns as compared to original approach. In a similar manner, Network Optimization, Fraud Detection and other Social Sentiment Analysis solutions were acceptable with lower confidence scores.
With the falling ARPUs, Operators’ need to optimize and reduce the wastage of resources is increasing day by day. Every dollar saved impacts the bottom line. Dynamic control of network resources is very much needed and Operators are getting this with the advent of NFV, SDN and other cloud based technologies. Instead of regular and proprietary telecom equipment / hardware, Operators are moving towards VNFs which can be deployed on COTS hardware. These VNFs can be powered up or down as per the network requirements. This impacts the energy costs directly and helps Operators save money too. This basic idea is further standardized in the 5G specs as well. The problem, therefore, is reduced to — deciding when to switch off or switch on the extra resources? How to decode traffic patterns? How to route the traffic among available resources? All these questions are answered by Analytics for telecommunication industry.
Traditionally, Operators knew that load increases during Christmas and New Years or during festivals. They also had an idea about the peak and off-peak hours during the day. Armed with this basic information, one could think that they are ready to cater to customer requirements and save costs too. As easy this looks, it is not easy at all to follow. The closer one can follow the load patterns, better saving can be achieved. Hence, to achieve finer control, one needs to have accurate data analytics function as compared to coarsely granular system of prediction. The question is not of “navigating” through the crisis, instead it is more of “surviving” through these difficult times. The result, therefore, is to have good analytical systems that can give clear outcomes with high level of confidence to make a decision.
While Network Management and Customer Quality of Experience enhancement is important, it is equally important to have proper Revenue Assurance System in place. This ensures that there are no losses happening in the network that go undetected. Traditionally, Revenue Assurance is being carried out by the Operator themselves or through consultants. Again, the accuracy of the analysis depends on the source of data and its quality. If they system under evaluation itself is the provider of the data for testing, it cannot be trusted completely. The outcome of analytics process will be biased. Many a times, a small error, negligible at smaller scale, can amplify itself to big proportions when extrapolated.
For any Data Analytics solution to have a good outcome with high confidence level, it is imperative that input data is as clean, granular and accurate (quality of data) as possible, discounting for the analytical process / algorithm / model inefficiencies. All the above withstanding, the need for good quality input data and thus high accuracy cannot be highlighted more. The Operators certainly need a good data capturing system which can capture data at such high speeds, an analytics system that can process the information with at high speeds and desired accuracy levels and generates insights on the fly.