In 2013, analytics professionals saw the hype around certain technologies continue to inflate. In some cases, the hype is real. For example, the marketing industry is embracing new methods of modeling data that have strong track records, and the results speak for themselves. However, other analytics technology may be more media hype than practical solution. Take a look at this list of the top stories from 2013 to find out how to separate hype from reality to make smart decisions heading into 2014.
Decoding marketing firms' adoption of analytics tech
Marketing firms have been fast adopters of analytics technology in recent years. For a business discipline that hasn't traditionally had much to do with technology, advertising companies are picking up analytics tools in large numbers. It's easy to see why, with the fragmentation of media channels and dwindling audience attention spans. Marketers are leveraging analytics tools to find the media outlets that potential customers pay attention to, and then delivering relevant, well-timed messages. To accomplish this, they are gathering much deeper data than in the past, as traditional demographic data may not be sufficient for building truly predictive models. Read more about why marketers are interested in analytics.
How to let data drive decision making
Few things can prevent analytics technology from delivering on its potential quite like executive objections. Studies have consistently shown that data-driven decisions outperform expert opinions. Yet at many companies, executives still make decisions based largely on their own gut instinct, even when the company has invested in sophisticated analytics systems. Ignoring analytics results is wasteful in two ways. First, it renders the technology investment meaningless. Second, it puts the company at risk for making less-than-optimal decisions. In this article, researchers and business leaders who have dealt with this problem explain how to get executives out of the way in order to embrace truly data-driven decision making.
No room for 'big data' in predictive models
The term big data has been around for years now, but precise definitions remain elusive. This lack of clarity over what the term truly means has fed wild speculation about what the concept might be able to help businesses achieve. One area where business users feel big data has limited applicability is in the development of predictive models. Building models is all about using samples from data sets to identify meaningful correlations within a sea of seemingly unrelated data points. But some feel examining entire data sets only amplifies the noise, making it harder to single out meaningful connections. Technology advancements have made it possible to store and analyze incredibly large samples of data. Some users can now feed their entire databases into modeling technology, but that may not always deliver the best results. See why some modelers say bigger is not always better.
The emergence of uplift modeling
The year 2012 was good for data geeks. One of their own, Nate Silver, who at the time blogged for The New York Times, became a star by correctly predicting the outcome of presidential balloting in all 50 states. But behind the scenes, a data scientist scored a much bigger victory. Daniel Porter, director of statistical modeling for the Obama for America 2012 campaign, helped propel the president to a second term. He used a little-known statistical method known as uplift modeling, which helped the campaign reach out to only the most persuadable voters. Now Porter is evangelizing the technique, and it is starting to catch on outside of politics. Marketers see potential in identifying the people who are most likely to need a nudge to buy their products.
Hadoop can't do it all
Hadoop was one of the most hyped pieces of technology in 2013 and practically became synonymous with big data. But businesses that are evaluating Hadoop need to separate the facts from the hype. It is extremely powerful and can be useful in many cases. But experts say users should understand the analytics tool's limitations and know when it is inappropriate. The main drawbacks users see are its inability to handle streaming data for real-time analytics and the fact that it still cannot replace traditional enterprise data warehouses for storing data. While solutions to these problems may be coming in the next year or two, Hadoop still has trouble living up to its hype.
This was first published in December 2013