Due to digitization, companies currently have vast amounts of data – Big Data – available to them. These come from very many sources. Social media, discussion forums, wikis or ratings communities on the Internet are just as much a part of this as company-internal databases or sensor readings. The targeted and comprehensive analysis of this data using statistical and quantitative methods and models has powerful business potential – for business decisions, innovation management, product development, marketing, customer relationship management and internal knowledge management.
Many companies therefore set up so-called "Analytics Labs", i.e. units that are separate from day-to-day operations and the line management organization. According to Harvard scientist Thomas H. Davenport, anyone who works there has the "sexiest job in the 21st Century": Data Scientist. But are Analytics Labs and Data Scientists really able to pick out from the flood of data precisely those results and recommendations for action that will help companies to work more efficiently and effectively? Not necessarily!
The added value of a data analysis depends significantly on the skills of the analytics team. In-depth understanding of the methodologies for analyzing big data needs to go hand-in-hand with comprehensive knowledge of the application context and of your own business. The focus on highly specialized methodology experts with a background in mathematics, statistics or physics, known as quantitative analysts or "quants", can be risky. Because no matter how sophisticated the mathematical methods and models may be, if you feed them incorrect, inadequate or poor quality data then they will follow the GIGO or "garbage in, garbage out" principle, leading inevitably to false conclusions and business decisions that do more harm than good. The same applies when methods and models used by Data Scientists do not properly reflect reality and are based on critical assumptions. A prominent example of this is the financial crisis. The assumptions that were used just did not apply.