Imagine a marketing campaign so successful that your offer sent to one customer reaches 361? Or the percentage of customers who decide to buy a service after a free trial is as high as 99 per cent! As proven in Grameenphone, this is possible, with the help of the big data and GP’s state-of-the-art campaign capabilities. By analysing a number of historical data created in customers’ interactions, it is possible to predict which customers are most likely to adopt certain services. Add to it the mighty and many times proven effect of viral spreading in the social group (“I want what my friend has” effect), it is actually possible to reach a much higher number of customers, giving them the services they want without even getting in touch with them.
“The customers like it – they are not spammed by too many offerings, but instead get suggested exactly what they want,” says Johannes Bjelland.
Together with his colleagues he has been analysing data for many years, still equally excited about how powerful data are, if they are just used.
Explosion of data
The volume of data in general is increasing – within one year we now generate the amount of data that equals the volume created during the whole last decade of the 20th century. And that was just a little more than 10 years ago. Telenor is a large contributor, just like Twitter, YouTube, Google and others. In DTAC only, there are hundreds of millions of voice and sms sessions collected every day. Every session creates additional data about caller, receiver, location, etc. All Telenor Business Units collect and store call data, in addition to product information, network performance data and many other details. By connecting all the data and using the proper tools and data mining techniques, we can get a deep understanding of customers, their habits and needs.
What is big enough data?
“The use of big data is developing. From just measuring what happened to understanding why it happens and ultimately being able to predict what will happen,” Pål Sundsøy says. He says that how much the potential of the data will be used depends on the corporate culture.
“With so much data giving you a number of accurate facts as relevant input for business decisions, it is no longer defendable to make decisions based on “gut-feeling” alone. Every business decision should be supported by insight from relevant data,” Pål says.
That’s why the team changed the buzz word “big data” to “big enough data”.
“The size in petabytes does not matter, but the impact the data can have on business decisions,” Pål says.
Not the black box
How are all these data chewed down to meaningful facts? It’s not like a bread machine where you put ingredients in and take a baked bread out and everything else happens on its own, or not a black box, as Pål and Johannes like to say.
The data mining process consists of work on data preparation – deriving the good variables, verifying data quality, and correct data errors, which is some 80 percent of work. All Data Mining is done by specialized software and special ‘machine learning’ algorithms.
Privacy is protected
It is important to point out that all data are anonymised, and that the researchers never see the names or the real phone numbers of the persons when doing the analysis.
“Our work is well aligned with local laws related to privacy and our internal policies that are even more rigorous, because no research can justify jeopardizing customers’ trust. In that sense, the customers and their privacy are well protected,” Johannes explains.