Data mining and telecommunication
How much can companies in the telecommunications industry benefit from big data and advanced analytics? It’s a strategic question. Every operator is looking for new ways to increase profits during a time of stagnant growth in the industry. As market is very mature, competition is fierce, especially as OTTs move into core sources of revenue. In addition, despite the fact that customer acquisition costs are high and margins are slim, capital costs are tremendous. It all pushes operators to search, research and apply innovative ways of increase revenues, and reduce costs.
Introduction
From one side there are a lot of data mining solutions in telecom industry which commonly used and applied to reduce customer churn, decrease bad debt, improve marketing effectiveness and customer experiences, identify network faults.
From other side in this highly competitive market operators need to make in addition research insights with big data to avoid the usual top-down approach, which sets up a business problem to be solved and then seeks for the data that might solve it. Usual way definitely have benefits, but it is unlikely to lead to any surprising results differ you from competitors — and it is difficult to execute until a company has demonstrated mastership in its use of data. Therefore, operators should start with the data itself, experimenting with what they have on hand to see what kinds of interconnections and correlations it reveals. This process must be started quickly and iteratively. If it’s done right, it forms the basis for more efficient operations and more effective marketing. At its best, this bottom-up method can give operators a more complete, transparent view of customers, enabling new and more profitable ways of capturing and retaining them.
This paper shows our approach to data mining process, how data mining can be applied in telecom for both commonly used applications and pilot insights to uncover the knowledge hidden within data sets.
…In the highly competitive market operators need to make in addition research insights with big data to avoid the usual top-down approach, which sets up a business problem to be solved and then seeks for the data that might solve it.
Process
Our data mining project lifecycle consists of 6 stages. It is nearly always require to move back and forth between different stages. And these movements depends on the outcome of each stage. The main stages are as follows:
- Business Understanding: This stage focuses on requirements from a business perspective and understanding the project objectives. Then it converts the gain information into a data mining problem definition and a preliminary plan designed to achieve the objectives.
- Data Understanding: It starts with an initial data collection and desription, to get familiar with the data, to identify data quality issues, to find first insights into the data or to descover interesting subsets to form hypotheses for hidden information.
- Data Preparation: It covers all activities like data cleansing, transformation and aggregation to construct the final dataset from the initial raw data.
- Modeling: In this stage we use diverse modeling techniques are applied and their parameters are calibrated to optimal values.
- Evaluation: In this stage the model is reviewed and thoroughly evaluated. We execute the steps to construct the model to be confident it best achieves the business objectives. At the end of the stage a decision on the use of the data mining results will be reached.
- Deployment: The aim of the model is to increase knowledge of the data. And the gained knowledge will need to be organized and presented in a way customer can use it. Based on it the deployment stage can be as simple as generating a report or as complex as implementing a repeatable data mining process across the whole enterprise.
Application in telecom
1 Reduce customer churn
It cost ten times and more to attract a new customer than to retain an existing one. This is where data mining is useful for the purposes of churn prediction. The key question is: How to predict the customers who are likely going to leave? Data mining techniques can be used to answer this question. By using data mining it is possible to generate the customer list with high probability to leave the company. Data mining techniques can help telecommunications companies to identify churn behavior patterns before the customers are being caught by more attractive offers from competitors. And plan the actions for risk customers to prevent the churn in advance.
2 Decrease bad debt
Data mining has been widely applied to make prediction for finance bad debt risk. In telecom it could be applied when shifting subscribers from prepaid to postpaid payment method or when selling in credit equipment like phone, tablets, modem and other. Application scoring models could significantly increase payments of services and products provided by instalments.
3 Tailor marketing campaigns to individual customers, develop new products and services
Relevant data mining techniques will help marketers’ with the right information which is behavioral market inclinations, change in trends, customers preferences, lifestyles, purchase decisions, and purchase timings to roll out customized products and campaigns at the right time on the right segment in order to capitalize market opportunities.
4 Identify network faults
Telecommunication networks are highly complex configurations of hardware and software and contain thousands of components, which are interconnected. These components are generating error and status alerts, which lead to a huge amount of network data. As these networks became very complex, management systems were developed to handle the alerts generated by the network elements. However, fault identification can be rather difficult because a single fault may result in a row of alarms—many of which are not associated with the root cause of a problem. Therefore, an important part of fault identification is alarm correlation, which enables multiple alarms to be recognized as being related to a single fault.
5 Implement value-based network capacity planning
To improve network performance and reduce capital and operating expenditures more effectively, operators need an approach of data gathering which facilitate a in-depth understanding of the network from a customer point of view, including services consumption behavior, his loyalty and spending. Operators need to understand, for instance, how many new customers sign up when they install a new antenna or enhance coverage in a given area, or how many existing customers upgrade their data packages when their network speed increase significantly.
6 Big data can even open up new sources of revenue.
Telecommunication companies maintain an enormous amount of information about their customers. No need to wait when some of you competitors discover knowledge for new revenue opportuneness. These opportunities could be related to some hidden economic issues or behavioral specific. Or could be opportunities such as selling insights about customers to third parties. Not to miss this you should start to explore and bottom-up research of your data as quick as possible and not to be afraid to make experiments and pilots.
…The possible goal of big data is to union and correlate every data source to create a holistic, clear, end-to-end view of all the cooperations every customer has with the operator.
Pilots big data researches
The nature of the bottom-up discovery approach is in gathering together all the data available to the operator, from internal and external resources; applying data analytics tools to process, analyze, and make sense of it; and finally defining how the results could be used. The point is to allow the data to “express itself”, finding out not just the obvious correlations and connections, but the unsuspected ones as well. Many types of data are available to operators — but it is unlikely that operators have all these sources at this stage — and certain sets of data might be combined to open up new business opportunities in areas such as campaign marketing and new products launches.
The possible goal of big data is to union and correlate every data source to create a holistic, clear, end-to-end view of all the cooperations every customer has with the operator. But to really enforce use of big data, operators must cardinally change the way how they gather, check, learn from, and make use of the data they have. Operators must learn from companies such as Amazon and Google, where data is lord and every product decision comes from what the available data says about customers and how it can be used. The big-data pilot programs should consist of teams of people from all over the company like network management, IT, marketing, finance, and sometimes even customers — who can bring their specific expertise to analyzing the data in new and different ways. They must know what it means to “play around” with the data, testing various combinations and correlations to see what works and what doesn’t. This process must be agile, repeating, and quick. Piloting teams need to conduct various tests on the data, learn from their mistakes and false starts, and move to the next experiment. They must avoid the overly structured mind-set that can let the pilot programs lasts for months and years. And they must speed up the evolutional process of discover and development, allowing the most valuable results to show up quickly.
Application examples
Verizon has found many use cases where advanced analytics have provided quite precise churn prediction in their mobile space within 1-2% margin. Imagine that this is a $100+ billion company. Overviewing the customer journey and combining those insights with the data warehouse, Verizon discovered that 90% of customers are using their tablet on the Verizon network in the first month of the contract, and 10% not. In month or two what is going to happen to these 10%? What happened to the 90%? Verizon divided them and identified what likely will happen in 3,6,9,12,18 months from now and advised the business whether they have a high churn risk coming. They also did the researches and learn to find the ways that could predict and stimulate usage of these customers so they not to decide to leave the network.
When T-Mobile’s customers start using data services at a higher than normal rate (for example in roaming), T-Mobile sends sms notification to the subscriber with advise of options for managing this usage, including information about alternative plans suited for usage, and advises the subscriber to call the call centre or visit the T-Mobile website for more data.
As well T-mobile with help of advanced analytics has an effective solutions to evaluate the effectiveness of its digital advertising, which is being incorporated into day-to-day marketing decisions and processes. More importantly, the company is building a foundation on which to build a sustainable analytics program.
Vodafone announced that big data analytics project has helped them to analyze the customer behavior and identify customer groups although the project is at a very beggining. Real time triggers on customer’s events have enabled business to target customer with right offer at right time. Big data analytics help to find customers sentiments by analyzing social media data and his overall Vodafone experience using self-care channels data. Social media analytics let business team to monitor various topics discussed on social media.
Conclusion
The industry as a whole spends far less on R&D than any other technology-oriented industry, and it’s the way where hidden opportunities still exist for every operator. Big data demands of every industry a very different and unconventional approach to business development. The operators that can incorporate new agile strategies and fearless experiments into their organizational DNA will get a real competitive advantage with competitors.
Big data helps telecom operators with real possibilities to gain a much more complete picture of their operations and their customers and to further their innovation efforts.