Optimizing customer service with machine learning

@ Orange

Key achievements

  • Built three Machine Learning projects up to production phase
  • Automated human-unfriendly tasks to facilitate customer service monitoring and improve marketing campaigns


Orange is the leading or second telecom company in most of Europeans countries as well as in Africa with a total of 266M clients.

I worked for Orange's French B2B data service for two years (2016-2018) :


With the building of the new B2B Big Data platform, the team in which I was working had opened up new opportunities : multiple sources of data about customer relationship were easily available in one place for the first time, as well as the tools needed to analyze them.

Customer service involve many repetitive tasks that can be automated and voluminous and various data, difficult for a human to interpret: Perfect cases for Machine Learning projects.


Building trust with business teams by predicting call reiteration to customer service

We needed first to find support among business teams to be untrusted with ML projects.

We imagined this supervised learning problem ourselves: if we could predict when a customer was likely to reiterate a call to customer service, actions could be taken to solve his problem and avoid dissatisfaction.

The first results were satifying enough to be untrusted with two others ML projects !


Facilitating customer service monitoring by automatically classifying reasons far calling

To fluidify the work of phone advisors, we were asked to automate the classification of the reason for calling based on a free text field.

We had to re-build and industialize this NLP supervised learning project.


Improving marketing campaigns by identifying hidden patterns in customer relationship

To target clients, the marketing team needed an automatic way to classify them based on their use of digital contact methods.

We built an unsupervised learning classification that met their needs.

I worked in collaboration with an expert Data Scientist that was only available one day per week.

I received guidance from him, mainly for data expertise and relationships with business teams, but I was the main contributor on all three projects.

Tools & methods :

Hadoop, Hive, Scala, Spark, Spark ML, Spark MLlib, Supervised learning, Semi-supervised learning, Unsupervised learning, NLP, Aglile methodology.

Interested in cooperation or would like to discuss anything ?