Customer Analytics - Improving customer satisfaction by analyzing millions of interactions

Context

As a Telecom giant with more than 250 million customers worldwide, a commitment for uninterrupted and ubiquitous service, cost and quality of customer support has become strategic and paramount for our Client. In order to maximize both user satisfaction and cost-efficiency, Our Client realized she could leverage its large databases of contact history to deliver support in a format best suited to each user, identify churn-critical areas or optimize CRM resources. Customer interactions number in the millions per day. They happen on a wide range of motives and channels, from phone call assistance to in-store repair to the increasing diversity of digital touch points. Our Client chose once again MFG Labs’ technical expertise to help them in the challenging task of mining complex and huge datasets into valuable strategic insights.

Solution

The magnitude of the challenge led to a breakdown in three main tasks:

1. Aggregating a huge number of data sources located in different infrastructures: each interaction channel comes with different technical tools to collect data. The goal is to centralize all relevant data in a HDFS cluster through a collection of ETL jobs.
2. Standardizing information in heterogeneous data: this step aims at building a common language to relate information from different sources. This is the sensitive crossroads between the technical, business and strategic viewpoints. It requires a deep knowledge of the business behind the data in order to understand how to extract a particular information from it. It requires also to define what information is valuable enough to be retrieved from all sources.
3. Providing on-demand analytics: now we leverage the previous work of data construction to address business-oriented subjects. How to identify unsatisfied customers? What is the actual impact of a new process/channel for customer support on CRM costs? How to dispatch incoming contacts into various channels/entities, depending on user profile and business needs?

MFG Labs put its technical and scientific knowledge at the service of these three purposes with a strong focus on the latter two.

We acted as a provider of on-demand analytics : for several known customer issues, we studied how millions of customers chose to interact with our Client. This analysis gave precious insights on how to define an optimal support process for each issue.

As for extracting standardized information from data, we worked on web page tagging: we designed a NLP algorithm that can automatically deduce semantic meaning from url text.

As a consequence, all pages from our client’s website are being automatically categorized by visitor’s intention. Each page is categorized within one or more categories. According to the sequence of pages visited, it’s therefore possible to understand the intent of each visitor.

With an active monitoring of these pages, crossed with CRM data, our Client can now detect at scale how to improve the website information. Our Client can also automatically insert valuable insights about the client’s intent in the CRM to help customer service operators.

Results

Our studies highlighted new strategic KPIs that have been included by our client’s teams in their business reports and monitoring.

The proof of concept, automatic categorization of the customer intent based on his journey on the website, helps our Client detect at scale sources of improvement, and potential unsatisfied customers.