Telecom data monetization¶
Real-time ads based on customer profile and internet browsing behavior¶
By: Younes Soliman
For the last decade, with the rise of social apps and cheap communication channels like VoIP, communications services providers were experiencing a hard challenge to keep up the level of revenues while knowing the fact that the revenues generated by their usual streams is diminishing.
Most of telecom revenues are coming from local, international and roaming services of Voice calls, SMS, value added services, internet subscriptions and usages, and some other services.
The providers were hit twice by the same hummer either going down or way up. That's because if they enhanced their internet services coverage, speed, and rates and put investments on the network, they will be paying all this, while customers will use more internet and decrease the dependence on old services streams. On the other side, if they don't enhance their internet services, customers will move into having the services from other providers and revenues will still decrease.
Here comes the idea of depending on the technology evolution of distributed environments in conjunction with machine learning techniques to release unusual services based on data and known as data monetization services.
Data monetization is a very wide area limited by the creativity of the business owner, data availability, and few times by technology.
Building a data monetization use case will also hit a goal on the path of digital transformation strategy. I will be talking here with little details about one of the use cases which is real-time advertisement implementation based on customer profiling and behavior analysis. When talking use case, we have to determine mainly 6 elements:
Data sources: From where we have to collect needed data. Are they real-time or batches, what capability are they able to handle, etc.
Analytical techniques: What information we have to extract from above collected data. Is the info trivial or will it need sophisticated methods and algorithms and what are those methods
Output: What we need to have at the end of data preparation, handling and transformation in order to feed other target systems. In addition how this output will be consumed, and through which APIs and integration techniques
Target systems: The platforms that will digest the feed information and emit actions.
Rules: What settings will be configured to ensure a condition happens?
Actions: The appropriate initiatives that will be launched once a rule is met
As known, telecom data is rich in information about customer details like demographic, location, handset, spend value, services usage, payment behavior ...and much more. The data can be used to generate revenues from third business parties who would like to benefit from the information and systems implemented by the CSP. One of the applications for third parties is location based targeting or what’s known as Geo-fencing. That is when a customer of specific characteristics (demographics, spend, etc.) and specific movement behavior reach at a specific point in place, he will be a target for the specific action.
The above use case is similarly done by rich data companies like Google, Facebook to target customers based on their search behavior, browsing behavior, networks, demographics, location, and other factors.
During proposal and implementation, the use case will face the ordinary two types of double "D" people on the positive and negative edges. Of course not where your mind took you but what’s meant are The "Double D" of "Don't Do" on the negative side while the "Double D" of "Day Dreamers" on the positive side. CSP should implement such use cases with caution and determined milestones to reach to the end goal of having a new stream of B2B services.
So here are some basic details about the possible attributes that can be specified in the data source element:
Demographic attributes like
- Nationality
- Language
- Date of birth / Age
- Gender
- Marital status
- Address
Handset Related Attributes
- Handset Make
- Handset Model
- OS
- Handset value
- Release date
- Other features
Service related
- Postpaid / Prepaid
- Individual/ Corporate
- Value Segment
- Corporate segment
Internet Behavior
- Searches
- Browsed websites
- Used Apps
Location
- Current Location
- Top usage location
- Other features
Analytics techniques like classification, Association, Segmentation, Text mining, or other can be used to generate hidden attributes.
Later i may specify the details of analytics models that can be built on top of data, what outputs that can be generated and what actions should be performed.