It is estimated that by 2016, 61 percent of web traffic will come from wireless devices as opposed to desktops, making significant contributions to the growth of big data. As customers continue to rely heavily on mobile phones to go online, the device becomes indispensable for multiple parties. Everyone from mobile operators to application developers is using mobile data to understand and reach their target markets. But how does this impact the mobile operators in the way they operate and interact with customers?
Traditionally, mobile operators' marketing teams activate from 10 to 100 campaigns per month in an attempt to verify or falsify a number of hypotheses in evaluating different marketing activities. With proper infrastructure, mobile operators can go to tens of thousands of possible interactions with customers - on a scale of a 2 to 3 order of magnitude increase - in their experimental capacity, according to chief data scientist Oliver Downs.
Operators traditionally viewed the Call Detail Records (CDRs) as the most detailed data for generating insights. However, there is a need to move to a sub-atomic level in order to study all sorts of customer interactions with the operator - usage, response to campaigns, purchase of new mobile devices, etc. Downs believes that expanding that level of granularity will allow the operator to generate a marketing campaign as quickly as the customers react, shortening the marketing improvement cycle.
Big data driven marketing is still at an exploration phase for many mobile operators. It requires deep integration with the IT team, which defines operator data according to its "temperature." Data accessed more than 90 percent of the time is considered "hot data", while data accessed only 10 percent is described as "cold data". The whole range from "hot" to "cold" defines the temperature of data.
T-Mobile's IT team suggests that operators define a strategy for managing multi-temperature data at different price-performance characteristics. Hot data needs to be stored at fast access storage systems while cold data should go to economic cost per terra-byte storage systems. The issue here is to get this done in an automated way without human intervention
With smart data management, the IT infrastructure is equipped to support all sorts of queries and data manipulation technologies (SQL, noSQL and/or map reduce). Big Data requires parallel access to data in all forms, including social graph, text, hierarchical data, and square data queries. The core technology capability that needs to get realised here is the "late binding" of data to the query time instead of binding data at the loading time.
A course of action that mobile operators may take that isn't easily realised through SQL, is building a whole big social network graph for their current subscribers to determine the most important subscribers; not from a revenue point of view but from an influence point of view. This type of network is illustrated in the following figure:
Another example of non-traditional use of queries is illustrated in T-Mobile's calculation of "Customer Effort." Their big data infrastructure is used to extract sessions of all customer interactions with every single department - call center, technical support, etc. - to calculate the customer effort and generate an n-path Sankey visualisation where each event and it's n-subsequent events are visualised to spot the different paths customers take when subscribing to a new plan or buying a new handset. The Sankey diagram below is a simple illustration of how you can trace the proportion of customers that gets directed from the sales shop to the call-center and then forwarded to operation support. The idea is to keep going forward with the flow (n-steps).
At TA Telecom we developed an analytic framework named Pi© (Performance Index) that analyses operator transaction logs and produces performance measurements for customer voice and data packages, as well as infotainment services. Those performance measures enhance marketing ROI, calculate opportunity cost and estimate the potential incremental revenues to help the operator decide which packages and services to communicate.
Pi also provides decision makers with sweet spots of meta-data associated with different packages. For example, a sweet spot can be discovered in customers between 20 to 25 years old that would yield the highest retention if subscribed to the 300MB data package and not a higher megabyte package or less. Pi uncovers actionable data to sales and marketing decision makers, not just insights.
More and more analytics on Big Data are evolving rapidly. Mobile operators all over the world are geared up for the potential of Big Data and Analytics to develop novel business ideas that have the potential of promoting revenue growth, optimising business strategy and reducing operational costs.
