EMEA Vision Award for Business Impact 2012

Timi wins the international competition EMEA Vision Award for Business Impact at the BI-Day 2012. The rules are: “The software solution that provides the highest substantiated business impact and the highest strategic advantage wins“. More precisely, the judges choose the winning BI-projet with respects to the following criterias:

• Is the solution groundbreaking (i.e., a new way to approach a problem, a transfer of information to a new industry, etc.)?
• What is the ratio of cost to return on investment?
• What was the impact on the business or organization?
• Was the solution adopted throughout the organization?
• Will the solution make it easier for the company to deploy future BI solutions?

We made a joined presentation with Real Impact about a churn-retention solution for MTN South Africa. This churn solution involves Big-Data computations on a large social network formed by the phone calls to extract the social communities, social leaders, etc. (i.e. This particular type of computation is usually named “SNA” computations= “Social Network Analytics” computations. Some people also call it “Graph Mining”). Here are some numbers:

We computed (with the SNA modules from Anatella) many KPI’s (more than 2500) based on the social communities and on the consumptions habits of each MTN subscriber. Using these KPI’s, we deployed a very effective predictive model (built with Timi) that predicts if a customer will soon leave MTN South Affrica. Thanks to this early alert, MTN South Affrica has enough time to deploy a targeted offer to prevent the churn.

The implemented solutions (developped 100% using Timi modeler+Anatella) is prefectly actionable because the whole process (data extraction from binary ASN-1 CDR logs, data preparation, data cleaning, SNA metrics computation, modeling, scoring) is 100% automated and runs in a few hours (10 hours) with the latest&fastest version of Anatella. The previous solution (developped using SAS) was not actionable because the time required to build a very simple model (no SNA, only 200 simple variables) was around 3 months. This very long duration made the process pretty useless because all the subsribers that were about to churn 3 months ago are completly unreachable 3 months later because they left the MTN network a long time ago.

Typically, the business impact of churn models is very large. When manipulating a large population (here: 21 millions subscribers), the ROI is even larger. We estimated that the ROI of the Timi predictive model for churn is several millions euros per year (although MTN did not disclose any precise numbers).