Over the last two years, the increasing relative cost advantages of Flash memory have driven a new area of competitive advantage in analytics: the ability to deliver real-time and near-real-time analytics on Big Data (aka Fast Data) that comes from sensors, smartphones and other Internet-connected data sources.
This is particularly true as firms investigate the potential of the Internet of Things, physical products that communicate with the Internet and generate sensor-type data on performance, user satisfaction, problems and the like. According to a , owner of this site, 49 percent of respondents said IoT would be either "critical" or "important" to future growth.
The new Fast Data analytics apps tend to draw their data from IoT streams that (a) are already in a public cloud or (b) involve a firm's physical products and services. To get the most out of these, the savvy user should understand three key value-adds that Fast Data analytics architectures and applications typically bring.
Choose Right Tradeoff of Speed-to-Insight and Data Accuracy
The typical Fast Data app runs on an architecture that splits between the real-time "shallow" analytics of massive sensor data-stream real-time analytics-required initial processing, and later deep-dive analytics done by traditional Big Data apps and engines. However, savvy users will find that some Fast Data infrastructure allows a dynamic tradeoff between write-to-disk data consistency and less-real-time analysis. The result is that the Fast Data app on the IoT can achieve much deeper analysis with no meaningful loss of analytical validity.
Discover New and Changing Patterns in Sensor-driven Data
A recent MIT study of GE's experience with Fast Data analytics on its Internet "cloud" connecting oil/gas industry "things" started with the expectation that these would allow GE's oil/gas products to work more cost effectively and reliably. This was true, but GI then discovered that data patterns also distinguished between environments in a far more detailed way and surfaced cyclical patterns and new patterns rapidly enough to avoid many disasters.
Join Public Cloud and Internal IoT Data in More Integrated Fashion
The same MIT study showed that customers of GE's IoT analytics-enabled products, once they accepted their worth, demanded that GE start applying the same analytics to other oil/gas products from other companies. This type of integration requires, effectively, availability of this data in a non-firm "cloud."
Likewise, car companies seeking to detect accidents waiting to happen around the corner must draw their sensor data from public clouds that are already providing traffic data. The benefit of such a marriage is a far deeper analytical understanding of the customer's and product's environment that fosters both longer-term relationships and better real-time response to a customer's needs.
The GE experience also offers one best-in-class way of starting to develop IoT Fast-Data analytics: Focus on enhancing existing products at the start, and then fine tune the three value-adds above depending on customers' evolving needs.
Wayne Kernochan is the president of Infostructure Associates, an affiliate of Valley View Ventures that aims to identify ways for businesses to leverage information for innovation and competitive advantage. An IT industry analyst for 22 years, he has focused on analytics, databases, development tools and middleware, and ways to measure their effectiveness, such as TCO, ROI and agility measures. He has worked for firms such as Yankee Group, Aberdeen Group and Illuminata, and has helped craft marketing strategies based on competitive intelligence for vendors ranging from Progress Software to IBM.