There are a whole load of companies who claim to have big data programs because they track data from their websites, even though the data they collect doesn’t really fit into the 4 Vs - volume, variety, velocity, and veracity. It may be that they are collecting a huge amount of data, but realistically is there much variety in it? Sure, you can look at who is visiting your site, where they’re from, and how long they stay on certain pages, but it does this really satisfy the 4 V’s?
However, some companies are truly embracing the idea of taking data from a huge number of areas and using them in their analysis.
A prime example of this is JP Morgan, who in December 2016 utilized satellite data to help predict company performance in some of the stocks they invest in. Using data provided by Orbital Insights, a company that processes images of mall car parks, one of the bank’s analysts, Rod Hall, who is normally a bull on Apple shares claimed that the company wasn’t performing well in Q4 of 2016 because data from the satellites made it clear that fewer people were in car parks.
This turned out to be a relatively reliable indicator of Apple’s Q4 fortunes, with the company actually seeing its first annual revenue decline in 15 years. This trend was bucked in Q1 2017 as the company saw their best quarter ever, but the data over the time period studied was accurate. Hall said of the potential volatility in Apple’s share price:
‘Apple is the one we would be most worried about because about 30% of their unit volume for the iPhone comes through the U.S. in Q4 normally, and Apple does more iPhone business through retail as opposed to through e-commerce channels. We believe over 50% of the population base in the US is near an Apple Store location. If the consumer ends up not doing okay, that would be the one most at risk.’
However, as the principles of big data show, the idea that you can glean a holistic view of an issue from unvaried data from a single source is wrong, something that Hall also pointed out in this analysis:
‘For Apple, it is hard to directly correspond sales to the data, because the stores are inside of mall, so car traffic alone won't tell you definitely.’
Another company taking this varied data approach is Tesco, who as far back as 2013 were utilizing weather data to help save money and improve the availability of stock in their stores. Through looking at historical sales trends, then comparing them to sales of particular items, they managed to make $100m in savings every year compared with before the changes were implemented.
It is something that also factors into Rod Hall’s calculations as it allowed JP Morgan to also predict why there had been a fall in the number of cars in parking lots:
‘You can't just look at cars because there's weather involved, and e-commerce is doubtless a factor [in some diminution of parking-lot traffic]. People tend to stay away [from the mall] when weather is nice, and we have had a warm fall, so that might be part of it.’
Companies often see that they have a large amount of data available to them through traditional analytics means and believe that’s enough to create a decent data program, but the reality is quite different. Collecting the data you have at hand can only get you so far until you need to start to looking beyond what you have, and begin to ask questions of why people are behaving in a particular way. Can you explain why the sales of one product spikes at specific times? If not, it suggests you need to look deeper than you currently are, which means looking outside of the data you already have access to.