Autonomous data analytics will be the driver of business analytics in the future and will be seamlessly integrated into our lives. John Thuma, from Arcadia Data, spoke at Enterprise Data World 2019 Conference in Boston about self-driving analytics.
He started off the presentation by asking the question if analytics initiatives are at an ROI cross-roads. According to a Gartner report on Business Intelligence project failures, 70-80% of the enterprise BI initiatives are bound for failure. The reasons for this are complex technology stacks, communication issues, lack of self service and relying on the past.
Descriptive analytics are great but they only tell us what's happening and trends, but they don't tell us why or what to do next. The organizations are now just governing the analytics; they are not developing analytics. Also, the friction between the teams in Business, Data Science, and IT areas (Data Science Iron Triangle) doesn't help much with success in analytics initiatives.
Thuma discussed the need for a transition from passive to active analytics, which involves:
- Human limitations to tireless exploration
- Passive analytics to "always-on" analytics
- Perspective lock to continuous inspection
Autonomous analytics will help with the data/information finding its customer, not the other way around. Solutions like recommendation engines make analytics seek out its user or consumer.
Augmented analytics also aid with this goal. Augmented analytics are the next generation analytics capabilities that can automatically prepare and cleanse data, perform feature engineering, find key insights and hidden patterns. Automation expedites investigation across millions of variable combinations that would be too time-consuming for a human to do manually.
Other initiatives like smart cities can also be benefitted by autonomous analytics, with alerts being sent to emergency support services and law enforcement organizations. Analytics can also help in redirecting the traffic configuration and alerting the connected vehicles with any critical traffic situations.
Legacy analytics are complex & expensive in the areas of ETL/data movement and are resulting in higher TCO and time to market. Some of the constraints to realize autonomous analytics are scale (can your BI and analytics keep up with growing data volumes?), performance (is BI fast enough to keep up with business?), and cost (are you getting the value out of your investments?).
Search based BI techniques enables the customers to have a conversation with their data where the users can build their own search engines without moving the data, using Natural Language Processing (NLP) techniques. An ideal analytics solution should require zero data movement, should work on-premises & in the cloud, be able to search across the organizations' data lakes and should be easy to setup and customize.
Thuma concluded the discussion by suggesting data analytics professionals focus on the "Design of Things", not just the "Internet of Things". IoT is about operationalization, whereas the design is about discovering what to build. Without autonomous analytics, you don't know what's important and what to design and build.
