For many organizations, real-time analytics is a means to automate the response of the business to ongoing events. It relies on streamed data to augment existing data stores to provide an accurate and timely account of the current and evolving situation.
In fraud detection, financial institutions capture transactional changes in real-time and combine them with historic account usage to help capture fraud as it happens. Further, streamed data plays a key role in the success of real-time decisioning models that assess risk related to loan applications.
In healthcare, the ingestion of data from multiple sources can improve patient care decisions while driving down treatment costs. Real-time analytics helps providers gain insight into patient health as it happens. The constant streaming of changes, updates and metrics ensures a better overall response for preventive measures and treatment.
In the retail sector, aggressive competition has caused many to shift to a real time or near-real time model. They seek to provide relevant and personalized experiences, enhance customer retention, raise search relevancy, and provide upsell or cross-sell opportunities. Their analytics engines depend upon a reliable and accurate stream of changes to customer behavior and market conditions.
But the frequent inaccuracy of marketing communications showcases how off-target things can get. For example, the current raft of ads being served to me includes (Advil – I don’t take painkillers), ladies bathing suits, pickup trucks (I drive a compact), skincare products, and jewelry. None are even close to my areas of interest.
When I research an article, I’ll often access a dozen vendor sites at short notice. Algorithms flag me as a hot prospect. For the next week or two, ads rain in on me peddling their products. Sales reps call whenever I download a paper, despite making it clear on the form that I am a writer and not a potential buyer.
Missing Link in Data Streaming and Analytics
There is no simple answer on how to marry up ads with interest. But one area that merits attention is the comprehensiveness of the datasets being analyzed. A common failing is only analyzing very specific types of data, unstructured data, web searches and social media being among those used the most.
What is often neglected is data sitting in systems labelled as legacy, such as the mainframe and IBM i. Such systems continue to play a central role within enterprise IT. More than 2.5 billion business transactions run on the IBM mainframe each day. Similarly, more than 100,000 enterprises rely on IBM i. These systems are cloud-enabled, virtualized and operate some of the most demanding databases and applications on the planet. That’s why they are still heavily used by many in healthcare, financial services, government, telecom and other verticals.
Due to barriers related to coding, computer language, security and general accessibility, vast troves of data from these systems may be missing from the data streams being used in real-time analytics. Compatibility is a major barrier. Legacy data often doesn’t play well with open-source frameworks and data formats. Application designers typically lack experience with older platforms. Native connectivity to a downstream target is another issue, and the processing requirements of legacy data can slow streaming to a crawl. Further, delays in capturing, processing and transferring changes recorded in such databases into the analytics engine can cause some to question the value of doing so.
But workarounds are available. Precisely Connect (formerly Syncsort), for example, helps organizations build streaming data frameworks for real-time analytics that can incorporate mainframe data without formatting issues, performance hits or other bottlenecks. Adding streaming data pipelines from IBM i and mainframe repositories to analytics engines helps reduce the incidence of inaccuracy. Perhaps one day, it could even end the days of online ad misfires that daily assault my screen and inbox.