There are many who tend to pour scorn on the sudden promotion of the role of 'Data Scientist'. At the very least, they view it just one more bandwagon; suddenly anyone who's added a smattering of PowerPivot wizardry to their reports, or has a little experience with Hadoop, is a data scientist. Yet, it's a real pity because the skill of the true data scientist is increasingly valuable.
The role of Data Scientist involves a lot more than fancy 'visualizations', and has nothing to do with the froth of Big Data or NoSQL. Data Scientists should have dual qualifications; yes, they need to be able to manipulate and report on data, but at the same time to be able to draw reliable conclusions. They must extrapolate from data, understand and be able to test for data quality, understand probability, cause-and-effect, variance, and advanced parametric statistics. Part DBA, part BI analyst, part statistician.
Business Intelligence must be more science than soothsaying. After many spectacular failures of database reporting, over the years, to be able to predict trends and provide what the marketing men term 'actionable insights', more companies are realizing that business data is stochastic, like the social science data for which parametric statistics were first developed. Measuring business indicators isn't like using a ruler. You need to be sure that the results are 'statistically' significant, and that you're not betting the farm on a mere quirk in the data.
I'm not sure that we need a separate profession of Data Science, beyond a narrow specialism. I think that anyone doing serious business reporting and data analysis should ensure that their skills are broad enough to take in the task of ensuring that the quality of data is sufficient, and that it is scientifically-justifiable to draw the conclusions you are providing to the business. That means being literate in parametric statistics.