Introduce DataOps to the equation. “Automate as much as possible,” said Chris Bergh, CEO and head chef, DataKitchen. “Check quality at every step in the process. Monitor operations in real time. Minimize cycle time and reduce errors to virtually zero. When these methods are applied to data analytics, it’s called DataOps.” Robust data streams across all sources are key to creating a real-time enterprise, added Asanka Abeysinghe, chief technology evangelist at WSO2. “Developers can build simple streaming and event-driven applications based on information. As a result, companies will be able to integrate and automate the use of more data streams. Automation will become the tool for development teams to optimize processes yet still capture datapoints across the enterprise.”
Develop use cases. The first priority in moving to real time is identifying priority use cases, according to Nuce. “Many companies are focusing on enabling predictive analytics to ensure point-in-time availability for item fulfillment and for consumer engagement and loyalty. With access to the movement of products in real time, brands and retailers can better pinpoint what a consumer wants when they want it and create upselling and cross-selling opportunities.”
Evaluate the technology and data governance structure that will support use cases. “For example, IoT sensors are being used to monitor goods throughout the supply chain,” said Nuce. “Tracking systems report valuable data such as location, temperature, humidity, and insights into quality control and traceability. Real-time data from IoT sensors—coupled with the sharing of quality, standardized data across trading partners within a supply chain network—can lead to a seamless flow of product intelligence. As a result, there can be more swift collaboration between trading partners to deliver on the promise to the consumer.”
Build gradually. Rip-and-replace is not an option for many companies, as “millions have been invested in data storage and management solutions,” said Corcoran, who advises deploying “a streaming analytics platform that can work seamlessly across [a company’s] current data management environment.” To do this effectively, he suggests ensuring compatibility with the major cloud platforms and computing architectures, interoperability with popular programming languages, and flexibility in terms of deployment method—on-premise, cloud, or hybrid.”
Remember, data quality is an essential element to any move to real time. “For those just starting their journey, it’s important to know that a prerequisite to real-time fulfillment is putting in the work to improve data quality through standardization and structure,” said Nuce. “Consumers need to be able to trust product listings and have the full picture of what they are getting, especially if they are unable to come into stores to see for themselves. This takes a commitment to global data standards and a dedicated data governance team that understands how data flows through technology.”
Look at TCO. Other factors to consider in the move to real time include total cost of ownership and the impact that any streaming analytics deployment will have on costs, Corcoran emphasized. “Look for a solution that has a low memory footprint and can run on commodity hardware. Companies should also consider ongoing maintenance and operational costs and what level of professional services are available to support the analysis, remediation and migration of data.”
“If there’s one thing we’ve learned over the years, it’s that the only constant is change,” said Bearman. “For example, the enterprise service bus was the king of real-time data delivery a decade ago, but today streaming technologies such as Apache Kafka have assumed the crown. Therefore, the key to building a real-time enterprise is to provide an architectural foundation that can quickly and easily adapt to new business requirements or technological innovations.” That means choosing real-time integration and data delivery solutions that can readily accommodate the addition of new data sources and targets, shifting data formats, and varying data delivery methods; all without relying on hordes of developers, said Bearman. In addition, he noted, these technologies should provide the ability to work on-premise, in the cloud, or in multi-cloud scenarios. “Finally,” he noted, “any real-time data delivery solution must have low-latency and cannot impact the performance of your existing applications or infrastructure.”