Scoring 2017 Conference: Eastwind Compares Industry Innovations with Its Own Products
30 august 2017
The annual inter-industry scoring convention finished in Moscow on July 13. The event gathered bankers, IT companies and mobile operators. Pavel Olifer, head of data analysis at Eastwind, was our man at the event and told us what new scoring trends had appeared in the presentations and how the new knowledge can be adapted for our products.
Pavel, what was your general impression of the conference?
“I liked the organization, and the quality of the audience and speakers. Forty topics were covered in one day. This is a lot of information: we discussed technical as well as legal subjects. There were people from the government, but mostly businesses. Naturally, the technical topics were the most interesting to me: what other companies are doing in terms of machine learning.”
So did you pick up anything new?
“First of all, I was pleased to hear we have captured the zeitgeist. I have compared the technologies presented by the speakers with how our Social Analytics product works: we are doing the same things, and somewhere we are ahead. For instance, director of Sberbank's risk modeling department Maxim Eryomenko explained how their analytic case system works – it is exactly how we do it. We also use Hadoop cluster and Python, and we also use Apache Spark to process data. He talked about the importance of regular assessment of scoring models – this is one of our focal points as well. Interestingly, even the industry giants have to cram it into the heads of their analysts: what has worked today may not work again tomorrow. In a nutshell, the tasks and challenges are the same for all. Eastwind is headed in the right direction. “
Did any other speakers stand out to you?
“Yevgeny Vinogradov from Yandex.Money spoke about anti-fraud scoring. There were many banks and financial organizations that shared their experience. Few have their own inhouse analysts like Sberbank does; most use scoring solutions offered by external IT vendors.”
Like Eastwind Social Analytics?
“Not exactly. Companies use standalone scoring software solutions, whereas Eastwind offers scoring as a small part of its large integrated platform. Social Analytics offers wider possibilities than just scoring. Eastwind’s product can be also used for collecting and deep analysis of big data, subscriber base anasysis, sample generation, and marketing process integration. At the same time, we have experience of creating scoring models based on our platform for telecom operators, banks and the retail. And we have new requests from customers. So, for the further development of Social Analytics we envision the possibility of creating a separate scoring module that could be supplied as a standalone product.”
Having been to the conference, which technologies should be a must for Eastwind’s scoring product?
“I, of course, could talk about the accuracy of analysis, support of multiple sources and technical innovations. But this is basically how it should be, and we are building all this into Social Analytics. So, based on what I took away from the conference, I would say a scoring product must be above all intuitive and friendly. There could be elaborated analytics inside, the work of tens of people. But for someone who runs the program it must be clear instantly how it works. With scoring, the most frequent question people ask is ‘why was I turned down?’ Or ‘why did they give me only so little money?' Lenders should be able to explain this and give some advice as to what could be changed to elicit a positive response. This is the main problem of the financial industry, since sometimes not even the lender’s officers can understand what has influenced the assessment results. Customers who were willing to pay off loans are frustrated; lenders lose profits. New generation scoring products have to fix this.”
More News
Pavel, what was your general impression of the conference?
“I liked the organization, and the quality of the audience and speakers. Forty topics were covered in one day. This is a lot of information: we discussed technical as well as legal subjects. There were people from the government, but mostly businesses. Naturally, the technical topics were the most interesting to me: what other companies are doing in terms of machine learning.”
So did you pick up anything new?
“First of all, I was pleased to hear we have captured the zeitgeist. I have compared the technologies presented by the speakers with how our Social Analytics product works: we are doing the same things, and somewhere we are ahead. For instance, director of Sberbank's risk modeling department Maxim Eryomenko explained how their analytic case system works – it is exactly how we do it. We also use Hadoop cluster and Python, and we also use Apache Spark to process data. He talked about the importance of regular assessment of scoring models – this is one of our focal points as well. Interestingly, even the industry giants have to cram it into the heads of their analysts: what has worked today may not work again tomorrow. In a nutshell, the tasks and challenges are the same for all. Eastwind is headed in the right direction. “
Did any other speakers stand out to you?
“Yevgeny Vinogradov from Yandex.Money spoke about anti-fraud scoring. There were many banks and financial organizations that shared their experience. Few have their own inhouse analysts like Sberbank does; most use scoring solutions offered by external IT vendors.”
Like Eastwind Social Analytics?
“Not exactly. Companies use standalone scoring software solutions, whereas Eastwind offers scoring as a small part of its large integrated platform. Social Analytics offers wider possibilities than just scoring. Eastwind’s product can be also used for collecting and deep analysis of big data, subscriber base anasysis, sample generation, and marketing process integration. At the same time, we have experience of creating scoring models based on our platform for telecom operators, banks and the retail. And we have new requests from customers. So, for the further development of Social Analytics we envision the possibility of creating a separate scoring module that could be supplied as a standalone product.”
Having been to the conference, which technologies should be a must for Eastwind’s scoring product?
“I, of course, could talk about the accuracy of analysis, support of multiple sources and technical innovations. But this is basically how it should be, and we are building all this into Social Analytics. So, based on what I took away from the conference, I would say a scoring product must be above all intuitive and friendly. There could be elaborated analytics inside, the work of tens of people. But for someone who runs the program it must be clear instantly how it works. With scoring, the most frequent question people ask is ‘why was I turned down?’ Or ‘why did they give me only so little money?' Lenders should be able to explain this and give some advice as to what could be changed to elicit a positive response. This is the main problem of the financial industry, since sometimes not even the lender’s officers can understand what has influenced the assessment results. Customers who were willing to pay off loans are frustrated; lenders lose profits. New generation scoring products have to fix this.”