Mario Mrljić is CTO and co-founder of Neos, a leading Croatian company with almost two decades of experience and number of successful domestic and international projects related to data warehouse, business intelligence, big data and data science projects. In 2015 Neos was recognised as world number 1 Oracle Specialized Partner in the Business Analytics category.
His primary interests include all business and technology aspects of data warehousing, business intelligence and data science projects as base for improving business decisions quality and helping companies to became data driven organisations. He has 18+ field experience based on a number of projects relating to financial institutions, telecommunications, leasing, retail and insurance.
Always keen to help customers on improving existing products/services and identify new business opportunities by using all available datasets and exploring new concepts and technologies in order to stay competitive.
Key ingredients of successful data science project?
Although we fully support the predictions advocated by most consulting firms that companies that do not invest in data science-based projects do not have a bright future – we are witnessing a large number of unsuccessful projects or ones with questionable profitability and expected results (according to recent Gartner research – big data projects failure rate is high as 60 and some argue that it even goes up to 85 percent).
According to data scientist practitioners and their leads there are few obstacles for successfully finishing data science projects. Often project starts from assumption based on previous experience or out of necessity to improve business operations using non-traditional methods. From assumption to hypothesis which need to be confirmed (or disproved) is a lot of unknown variables or at least not clearly defined ones. Project scope as well as duration, data sets and necessary tools are all subject to change as project develops. Additional specificity of data science project are project results. Final conclusions in form of structured presentation needs to properly and clearly expose project result. As results are not in form of application or report, it is expected to presentation material explain and assure contractors how project conclusion and developed modules will be integrated in business processes and IT system.
Since we at Neos do not like to participate in failed projects – and would like to continue with such practice – we decided to develop data science projects specific methodology/framework. This presentation will shortly explain methodology combining different concepts (ranging from design thinking to system engineering) which includes all steps, tools and expected results in context of mentioned data science project specifics with aim of project transparency and successfulness.