Analytics Magazine Analytics Magazine, January/February 2014 | Page 33

shown in Figure 3. This probably reflects the use of new data types to improve the predictive power of existing models. Teams are still beginning largely with traditional data types, but they see increased value from new data types once they have Figure 3: More experienced practitioners use more data types. some success. With more successful, more established stored back into a database for later use teams using big data more broadly, it to real-time scoring. This shift is reflected seems likely that there will soon be a in the increased use of intra-day and realrapid and significant growth in the use time data in predictive models. As one surof new data types in building predictive vey respondent put it: “Intra-day data will analytics. More traditional structured be the most valuable to our company since data will likely remain broadly central we are open 24 hours.” to effective predictive analytic modScoring streaming data is not yet a els. One survey respondent put it this mainstream use case though it seems way: “Big data is a misnomer as data likely that the general shift to a more eventhas always been big. The challenge centric, real-time world will bring it squarely is making use of semi-structured and into the mainstream before too long. unstructured data in solutions. This Back in 2011 it was clear that early will be the next giant leap forward in adopters were going to get an edge. using data.” They were more likely to have plans for The velocity of data also matters. Prebroader deployment and saw predictive dictive analytics is increasingly focused on analytics in the cloud as more valuable. near real-time, op erational data. This kind This trend strongly repeated in 2013. of data grew the most in importance beOnce again, early adopters with one or tween 2011 and 2013. This corresponds more use cases deployed were signifito the general shift in predictive analytcantly more likely to have plans to exics from batch scoring where results are pand deployment. Similarly those with A NA L Y T I C S J A N U A R Y / F E B R U A R Y 2 014 | 33