Last week, I had the privilege of taking a sneak peak of IBM Watson Analytics; a cloud-based service that aims to provide data analysis for business users. It is very new and the beta version is yet to come out (although you can sign up). IBM says Watson Analytics is their biggest analytics announcement in a decade! From the demo video shown at the IBM conference, a user typed in natural language questions and the dashboard displayed the relevant metrics and graphs. I would say the audience response was mixed partly because people tend of be skeptical about new technology by default! Everyone was very impressed by the Watson technology, but there were still some concerns. From the companies I spoke to, the concerns fell into three areas: approach, database integration and data governance. These are challenges not unique to Watson but for analytics and data science in general The 'deus ex machina' approach to Watson Analytics has been questioned by some industry experts. The senior director of data science and analytics at Samsung emphasised that by giving a blackbox to business users, it moves away from a true understanding of the data, and could lead to dramatic, and surprising failures. Watson Analytics has a neat feature where it provides the confidence level for the metrics which helps to address this potential problem. I also think that for many applications, it does not require users to have a thorough understanding of the data. As the industry progresses, the different applications will inevitably be better defined and understood. Analytics and data science is a vast field, we are only at the beginning and having more tools means companies have more options to choose something that really works well for them. Using the analytics platform requires uploading the dataset into Watson's column-based database cloud. I spoke to a few companies who said that it would be a huge challenge to find a way to transform their highly dispersed datasets into a form acceptable to Watson. This perhaps is the biggest hurdle that prevents wide adoption analytics. For many companies, especially large corporations, data engineering what is preparation of the data, big data and legacy systems are a real struggle. It is something we have to work hard as an industry to fix. Lastly, in relation to data governance, some audience members expressed that their dataset has to stay within the UK boarder. The cloud-based service, hence, could poses a problem for these businesses. IBM mentioned they would be/already have setting up a data centre for Watson Analytics in the UK to help solve this issue. Despite these challenges, there are many things to be excited about the platform. It would work very well for generic applications and propel data driven culture in businesses. It has greatly improved accessibility for business users. The user interface looks clean and easy to navigate. Running good predictive modelling using Watson Analytics is likely to still require highly technical staff, those with a good command of SPSS, and an understanding of the different machine learning models. To me, there seems to be no shortcut to it: in order to carry out real data science, you need to know data science. In any case, it is worth signing up to the beta version and get an early introduction.