2000px-Orange_logo.svg.png
0f7c5f6.jpg
“PredicSis Machine Learning is impressive – very fast plug ‘n’ play. I was particularly impressed by the speed of integration and I am looking forward to seeing the results of the forthcoming campaigns. For Orange Romania, these were the first tangible results we’d seen in the big data analytics new world.”
— Julien Ducarroz, CCO

Bad Debt reduction with Orange


The Client: Orange

Orange is the market leader in Romania with more than 10 million customers.  Orange offers a wide range of communications solutions to its clients, both individual users and companies, ranging from basic services to full service voice, data, fixed and mobile, and TV services.  Orange has the best 4G network in Romania – 96% of users certify 4G experience. In 2016, Orange became a convergent fixed-mobile operator, launching fibre optic broadband, fixed voice and digital cable television in 3 major cities, and in the latter part of the year it extended service coverage in urban areas.

The Challenge

Collecting payments is a vital day-to-day operation for a company.  Tight processes exist to ensure cash in, but unfortunately customers do not always pay on time and it’s difficult to push too hard without upsetting them. 

Which customers can be relied on to pay eventually? Which invoices should you adopt a more hard-headed approach with?  Orange Romania wanted to adapt its collections process by taking the most appropriate action for each invoice and each customer.

The Solution

PredicSis.AI allows Orange Romania to fully exploit every piece of data generated by the customer before the invoice is processed (CRM, Call Detail Records, payment history etc). In just a few clicks Business Analysts are able to deliver actionable insights to evaluate the financial risk of all the invoices at the individual customer level.

Constructing the Predictive Model

Past payment performance is reviewed regularly.  Thanks to the data sources describing customer behaviour, early signals of regular payment, late payment or non-payment are detected.  Combined, they form a predictive model able to distinguish and separate out bad payers from good payers.  This model is computed automatically.

Accurately identifying the risk of each invoice

Once the model is ready, it can be used to assign a risk level or ‘score’ to each new invoice. Then, depending on the score generated, the most appropriate recovery process is applied to the invoice - from a simple reminder message to immediate service suspension and intensive recovery.

The Results

PredicSis.ai has enabled Orange Romania to increase its recovery rate by more than 20% without increasing the overall recovery costs.  By optimising the intensive versus non intensive recovery mix, Orange Romania has also seen an improvement in customer satisfaction.  Finally, understanding the key

signals will also help to improve product design over the long term.