The social media buzz around the Data Science Enthusiast Meetup in Istanbul on 17th Feb was hard to miss. Tuba Islam, Senior Business Solution Manager at SAS, was one of the speakers and she hosted a session around Analytics in Action. I caught up with her after the event and interviewed her about the meetup and her work.
How was the Istanbul meetup different from the others you have been to?
Tuba Islam (TI) - I joined a couple of meetups on various technologies in the UK, which were very specific – focused on one single topic. They attracted mostly technical people. This meetup in Istanbul was more business oriented, and focused on how you can apply the technology on business. We shared lessons learned and tips, which was very useful. The other meetups that I’ve been to, were missing the application to business, which I find the most important part of the analytical process.
Another difference is that in the UK, there are many meetups, but in Turkey, this is quite new and we had a good number of attendees. We had three speakers: an industry expert Ismail Parsa, Chief Data Scientist of Hepsiburada.com with massive experience in Data Science, he just came from Amazon in the US; one person from a university, Prof. Dr. Burçin Bozkaya, Professor of Business Analytics at Sabancı University, who started a project with MIT: Big Data Lab and explained what they do in that project. I represented SAS focusing on applications of analytics in different business domains and what helps when you build a project. It was a good mix of sharing experiences and looking at Data Science from different perspectives. We had lots of students in the room, from engineering and statistics but also social skills and business administration students. It was quite intense discussion around many topics, and it was very insightful.
What were your key takeaways?
(TI) - Most of the students were trying to understand their options: what kind of career opportunities they might get with data science, what they can do when they graduate. They got a good understanding of where they can apply data science: not only in retail but also in banking, insurance, energy, and other industries.
Another key takeaway was that a lot of Data Scientists were looking for ways on how to handle deployment issues. If you build a perfect model, but you cannot deploy it quickly and do not have it in production in time, what is then the value of analytics? We all agreed that it is better to build a model that is good enough to deploy, rather than building a very complex model that is hard to implement. This was also emphasized by Ismail Parsa, Chief Data Scientist at Hepsiburada.com.
Focusing on the whole process of applying Data Science is important: start with the data, discovery being the next step and deployment is then crucial for success.
There were questions around the issues of using different modeling techniques and having to decide which model will perform the best. I explained how you tackle this with model comparison and out-of-sample and out-of-time validation.
And last but not least, we discussed the expectations set to a Data Scientist in the market, and how the role can be misinterpreted: people might think all you do is building datamarts, running queries or creating reports. The analytical culture in the organization is very important and if you are working in an analytically immature environment it can be very hard to convince people and make them use the analytical outcomes. You need to be patient. It just takes time but it happens. Once you show the value, you could then proceed to automate the analytical processes.
Given the important of this topic, what advice would you give to future meetup organizers?
(TI) - First search thoroughly what is out there and where the gaps are – there are so many meetups already. It’s important to do something you are really enthusiastic about: choose a topic that is close to your heart. It’s like growing vegetables: if you like carrots, grow carrots. If you don’t like peas, don’t do it. The audience can easily feel whether you are passionate about a topic or not.
Also, make sure you have a story to tell, it’s so much better to show how you can apply the technology rather than demonstrating just the technology itself. And of course make sure you have enthusiastic speakers.
How does this meetup tie into your normal work?
(TI) - I work in the global practice Analytics team at SAS, and I am involved in various projects from different industries. One day I work with an energy company to forecast half-hourly electricity consumption and the other day I work with a telecom company to build propensity models for their campaigns or analyze their customer complaints from contact center for churn detection, etc. I spend a lot of time with the customer, to prove the value of the software. In order to achieve this, you need to know both the technology and also the business processes. That’s why communication and collaboration is very important in data science. And that’s what I like most about my job. Meetup is a good opportunity to share this knowledge and connect with a wider audience.
Having this experience from different industries is very valuable and gives a bigger perspective on Data Science and Analytics, and you can recognize that the same approach: data – discovery – deployment - can be applied in all business cases. The actors might change, e.g. it could be the retail customer, the patient at the hospital, the merchandiser. However, the process stays the same. We call this Analytics in Action: not only gathering and analyzing data and building models, but also executing it in production very quickly, real-time or batch.
What changes are you seeing in the field of data science?
(TI) - It is now even more popular thanks to the recent advances in the technology – the improvements in the computation power, the affordable data management options. More people are getting involved and the market is growing, there are many niche and big players. SAS is investing in new creative solutions to support the new trends, providing a platform that supports the enterprise open ecosystem.
Collaboration will be more important: people with different skillsets working on different environments need to be able to work together. It’s important for organizations to focus on the bigger goal and provide the environment to work together easier. The automation will also be more crucial.
Cloud and the service-oriented approach will get bigger. Using tools and architecture in the cloud rather than having the applications in-house. With the service-oriented approach I mean initiatives like Analytics As A Service, where you receive the results and you don’t have the skills in-house. This is due to the gap that we currently see: Requirements are increasing faster than the resources. So there is a need for more service oriented approach.
One thing that will not change though, is the importance of deployment. Deployment is key for achieving and proving value from the data.