Organizations are excited about predictive analytics and machine learning for a number of reasons. Companies want to better understand customer behavior. They want to better predict failures in their infrastructure. The uses for predictive analytics are extensive and growing.
Fern Halper, Ph.D.
Enterprise analytics spans a wide array of categories but they all have one thing in common, they require human interaction to realize value. However, much of that value is often left on the table. Factors such as user interviews, persona design, stakeholder buy in, wireframing, iteration, adoption and feedback are underutilized and greatly increase the risk of user disengagement and stakeholder frustration. Analytics managers and dashboard creators can miss the opportunity to leverage user motivations to drive success.
Data science offers great potential for what it can contribute to business strategy and operations—that is, if data scientists are actually able to do data science rather than spend most of their time on data management and preparation. TDWI finds that most data science projects spend the majority of time on these areas rather than on development of analytics, models, and algorithms. To increase business value, organizations need solutions that will flip this ratio.
Gut instinct alone is not enough to enable decisions that will drive success. Most businesses today believe in the power of BI and analytics to help drive insight and value. TDWI research indicates that the vast majority of organizations are using technology such as visual analytics and BI dashboards to help them gain insight. However, gaining insight and using that insight to make decisions are often two different things.
Fern Halper, Ph.D., David Stodder
Hortonworks, Looker, SAP, SAS
A ten-fold increase in worldwide data by 2025 is one of many predictions about big data.
With such growth rates in data, the “data lake” is a very popular concept today. Everybody touts their platform capabilities for the data lake, and it is all about Apache Hadoop. With its proven cost-effective, highly scalable, and reliable means of storing vast data sets on cost-effective commodity hardware regardless of format, it seems to be the ideal analytics repository. However, the power of discovery that comes with the lack of a schema also creates a barrier for integrating well-understood transaction data that is more comfortably stored in a relational database. Rapidly changing data can quickly turn a data lake into a data swamp.
As Internet of Things (IoT) technologies become more common and web data grows in volume, there is growing evidence that the ability to analyze continuous data is not only valuable but necessary. In fact, those with the ability to capture and analyze massive numbers of independent continuous data streams will have a powerful capability that will help them to power operational intelligence and predictive analytics. A growing number of applications increasingly rely on fast analysis, but tomorrow’s world will be even more dependent on up-to-the-minute consumption of data streams.
Most organizations believe they will achieve better analytic results if they populate a deeper bench of experienced data scientists and machine learning practitioners. But this is akin to building a home exclusively with highly skilled framers, brick layers and cabinet makers. You’ll end up with a solid structure and great workmanship, but not a true functional home.