CaseStudy4

South Africa, Public Sector

MOBILE NETWORK CUSTOMER CHURN

Constructed a machine-learning model using off-the-shelf machine learning technology

48 hours

to build predictive machine learning model

6 hours

to train client team to self-sufficiently build and refine a model

48 hours

to build predictive machine learning model

6 hours

to train client team to self-sufficiently build and refine a model

$100

And under = tech costs

80%

Model accuracy (AUC =0.8)

$100

And under = tech costs

80%

Model accuracy (AUC =0.8)

THE CHALLENGE

A mobile operator was under pressure to identify the target customers of its marketing efforts in order to prevent churn. It expected a sudden loss of prepaid customers due to the market having 80%+ prepaid penetration and low barriers to switching operators, which resulted in customers rapidly switching to other mobile networks. The challenge was to minimise the impact of the company’s marketing campaigns which were inaccurately targeted at customers.

OUR APPROACH

  • Constructed a machine-learning model using off -the-shelf machine learning technology
  • Used the model to identify valuable data and predict customer behaviour
  • Designed a training programme to help the client team continue to use, refine and build other models of their own