Telecom operators have long sought the holy grail of how to drive increased customer loyalty and profitability. In an industry notorious for churn, being able to identify customer pain points and potential opportunities and drive improvements can be invaluable. Orange’s artificial intelligence (AI) and machine learning (ML)-powered Khiops solution can exploit big data sets to power that transformation.
Big data refers to data sets that are too large or too complex for traditional data processing applications to handle – and telecom operators are absolutely full of data. Call records, mobile usage, network equipment, server logs, billing, social networks, all provide information about customers and networks. How can telecoms operators use it to drive greater customer satisfaction, increase loyalty and improve their business?
Khiops is an AutoML solution designed specifically to help any company, including telecom operators, maximize their data and use it to drive business improvement. It is an AI and ML-powered tool that takes large data sets, analyzes them, creates actionable business insights, and implements them using automation.
For example, customer churn has always been one of the biggest business problems. According to Harvard Business Review, it costs between 5 times and 25 times as much to find a new customer than it does to retain an existing one. What if data could help you reduce that churn?
It’s really possible using a tool like Khiops. Operators have lots of data on each customer: their consumption, interactions with the company, the services they buy, all of which could be used to assess their satisfaction. It is normally the role of data scientists and business experts to identify which information is relevant to the question at hand, but that is usually a complex and time-consuming exercise.
Khiops has been built with simplicity in mind, and one of the tool’s key strengths is automation of feature engineering.
| Feature engineering is an essential part of data science protocols, which refers to the “process of selecting, manipulating, and transforming raw data into features that can be used to benefit the business using ML. A “feature” is any measurable input that can be used in a predictive model, for example the color of an object or the sound of someone’s voice. Regardless of data or architecture, a badly-generated feature will have a negative impact on your data model”. Source: Towards data science - What is Feature Engineering — Importance, Tools and Techniques for Machine Learning | by Harshil Patel | Towards Data Science |
Traditionally, feature engineering is a time and resource-consuming task. For each business challenge you want to address with data, a number of data features must be developed. Data scientists examine available data and develop those models, but it can take many days to do so.
Khiops automates this task by executing algorithms that, according to the task’s objective, such as which clients are more likely to churn, agnostically identify all the features relevant to remedying that issue. In other words, Khiops is able to identify the most important correlations between any feature and the objective. Moreover, by doing it quickly and efficiently, Khiops saves time and cost on expert resources.
According to Felipe Olmos, Data Scientist at Orange, “Khiops is the end product of a 20-year project for us. We have a dedicated team permanently-employed to build this market-leading tool that can help maximize insights from data and solve business challenges, such as preventing customer churn.”
Orange has developed Khiops from a strong scientific base. Its creators are data scientists and the core Khiops algorithms have been peer reviewed in scientific publications, to ensure the solution’s quality and to reassure users. This full transparency allows industry experts to evaluate the algorithms for themselves and understand how the tool works and provides such effective results.
“We believe that peer review shows our commitment to transparency and validates our work on Khiops. We are very pleased that Khiops is helping make data analysis more accessible.” says Olmos.
Orange has built up other example use cases where Khiops has demonstrated its value. It can be used for sentiment analysis, where comments online and on social media are gathered and analyzed to see if they are positive or negative.
Khiops has also been used to identify small offices, home offices (SoHo) in Africa and the Middle East, companies that are not registered anywhere and so can’t be reached out to make them targeted offers.
The tool can also play a vital role in helping carriers identify and block fraudulent voice calls, by automatically analyzing large databases of calls made and calculating probability rates that such calls are fraudulent. Human experts can then make informed, and rapid, decisions to block these types of calls. Khiops is actually one of the core components of the anti-fraud solution used to protect Orange wholesale customers’ international voice traffic.
To learn more about Khiops, please visit: https://khiops.com/
