Real-time Analytics


Real-time analyticsOrganisations are focusing more on real-time analytics to incorporate advanced analytics in their day-to-day decision-making processes. There are two very different approaches to implementing advanced analytics in real time.

The first approach is to apply ‘conventional’ advanced analytics as we know it to operational data sourced in real or near real time. The challenge with this approach is to source and prepare the data quickly enough in order for the analytical engine (like SAS or SPSS) to quickly produce new insights, based on very recent events or interactions.

With this approach you have to face the “data warehouse vs source system” decision. Although the data warehouse is always the better data source for advanced analytics, with the data being conformed, integrated as single source of the truth, quality checked, etc., most of the advanced analytics modeling engines do not require the data in a format typically provided by the data warehouse. Many of the advanced analytics modeling engines prefer data in flat file formats. Of course the second issue with this approach is whether the data warehouse is updated quickly enough after the events or transactions happen in the business. Organisations with a drive to implement real-time or near-time operational reporting from their data warehouses are usually in a better position to feed real-time analytics off their data warehouse.

The second approach is to apply data analysis to the data as it is streamed in, in real time. With this approach, a slightly lower class of analytics is performed; however it is performed in real time, on data streamed directly from its origin. With tools like SAP’s Aleri, the data flows “through” the queries in memory to perform classification, filtering, data cleansing and calculating some real-time insights. Used in conjunction with its sister product RAP, it can then compare the real time data to historical trends. As with any in-memory processing solution, the key issues that we have to cope with are volumes, velocity and complexity. It is also only business-effective, if the business can react or adapt to changes in trends in real time.

Leave a Reply