In the Era of Big Data, it’s easy to assume thatall large collections of data are ‘Big Data‘. For example, large manufacturing companies and warehousing facilities might have years’ worth of inventory data, many terabytes in fact, but this isn’t necessarily Big Data. Similarly, data from 1,500 PoS cash registers isn’t Big Data, nor is a large collection of spreadsheets.
These are examples of lots of data, and indeed, businesses need effective ways to store, analyze, and use these types of data. It just isn’t Big Data, and there is no need to invest in data lakes, data scientists, and a gaggle of Hadoop products to manage it. There are other ways to handle large sets of data. How can you tell if your problem is just managing a whole lot of data, or if your issues call for Big Data solutions?
1. Does the Data Come from Multiple Types of Sources?
Data professionals refer to the three V’s of Big Data (or the 4 V’s, depending on whom you ask). These are: Volume, Variety, Velocity, (and if you’relooking for #4) Veracity. The second one is what we’re talking about: variety. Big Data does not typically come from a single source or system (though there are exceptions). It usually involves data from numerous sources, that comes in a variety of formats, and includes a lot of variables. For example, your PoS data wouldn’t be big data, no matter its volume. But if you wish to integrate it with data from your suppliers to manage the supply chain, then it becomes Big Data. The complexity is what makes it Big Data, not the mere size.
2. Does the Data Need to be Analyzed in Real-Time?
One way that data can lack variety and still be considered Big Data is when ithas to be analyzed in real-time, such as for fraud prevention or for use in stock trading. Fraud prevention in the credit industry, for example, isn’t necessarily that complex at all. But it does require real-time analytics techniques (generally Spark, perhaps Hadoop and Spark) so that fraud can be detected and stopped instantly at the PoS. Similarly, stock traders depend on high-frequency trading data that also isn’t extraordinarily complex, but has to be processed instantly in order to make timely decisions to purchase or sell stocks.