[Editor’s Note: Click here to read Part 1 of this three-part article].
In the first part of this series, we went over the definitions of Big Data as they relate to a Broadband Service Provider. Now that we understand the kinds of data that such a provider may have access to, let’s look a little deeper into what can be done with it.
Obtaining Actionable Insights #
The aim of any Data Analysis effort should be to provide “actionable insights”. An actionable insight is an understanding (an “insight” obtained by professional analysis of data) that points to a course of action. Insights that help in understanding network status or customer behavior may be interesting, but are not useful to the Broadband Service Provider unless they are actionable. A specific action, beneficial to the provider in some way, can be executed based on the information generated.
Let’s take an example. Say a network provider has seen a slow but steady decline of Revenue Generating Units (RGUs). The Broadband Service Provider wants to reduce the churn of RGUs. How can Big Data help?
If the Service Provider were a small provider, it may be possible for them to talk to almost every customer and understand their frustrations, issues and maybe work with them to prevent RGU churn. However, there comes a point where it’s just not possible for a Provider to talk to every customer before they have approached the Provider to terminate the service – and that is where Big Data can help.
Data collection, machine learning and predictive analysis could together point to high risk customers who must be approached proactively, before the customers approach the Service Provider to cancel service.
It’s the difference between going to a physician after an ailment occurs (the ailment may be impossible to cure) and having the physician inform the patient before the ailment occurs that they have a very high likelihood of getting a certain ailment unless they change a very specific behavior immediately. We would prefer to find out before the ailment occurs (remember predictive analytics?) so that appropriate preventative action may be taken.
What if the Service Provider could predict that a specific customer (RGU) has a high likelihood of leaving, before the customer approaches the Provider to cancel service? Such knowledge could be potentially very valuable to the Provider, and would qualify as actionable insight. The insight was that a specific subset of their customers have a high likelihood of leaving, and the action possible is to contact them and attempt to win back their business.
Good Data – Data that is true unto itself #
Garbage In = Garbage Out. In order to get good results from performing analytics on data, we first need data that has high intrinsic quality. Inaccurate data will simply give us inaccurate results, and insights or predictions generated may not be dependable. What a Data Scientist needs is access to “Good Data”.
So, where do we get this data from, and how do we make sure that it is of high quality? Good data has at a minimum the following characteristic:
- It is data that is from a relevant time-frame (it is recent enough to be relevant today),
- And it exhibits consistent correlation (the data is true unto itself).
For a Service Provider, this data could be infrastructure, customer, social or even device data, as we discussed in a previous article. A qualified Data Scientist must work together with a domain expert to ensure good data is collected, analysed and acted upon.
A good practice for a Broadband Service Provider is to make sure that, until a full assessment of their systems and practices is performed by qualified experts in the Big Data space, they should plan on collecting as much data as is practical. It’s better to have too much data, than to regret not having collected enough information.
In the next part of this series, we will explore some scalable, reliable technologies to store this data and perform analysis on it with the aim of providing specific actionable insights.