These 4 Big Data Myths Could Be Hurting Your Business
March 31, 2017 • Blog Post • By Quartz Creative
IN THIS ARTICLE
- Ken Rudin, director of analytics at Facebook debunks four all-too-common misconceptions about analytics in business
- Rudin says that valuable big data insights are usually much more accessible than you think
Why bigger doesn't necessarily mean better
There's only one wrong way to think about Big Data, according to Ken Rudin, director of analytics at Facebook. In his keynote at the recent HP Big Data Conference in 2015, he told the audience it is inaccurate to believe everything about Big Data has to be big.
The buzz about data processing in the Cloud has convinced many CIOs that they need teams of math Ph.D.s and computer scientists running Hadoop when in fact, major insights are usually much more accessible than stakeholders realize. Here are four all-too-common misconceptions about analytics in business.
Myth 1: Big Data is putting huge quantities of information into Hadoop
There is no ideal quantity of data for analysis. Even if youve only just begun collecting information, you dont necessarily need to wait to begin using it. In fact, its rare that a company would process a massive data set all at once even when they have one. For most queries, you only need to access a small portion or sample size to begin deriving useful insights. Asking targeted questions usually produces more actionable insights, so in many cases, smaller, more segmented queries are better anyway. You can get these smaller-scale insights out of legacy relational databases without necessarily needing Hadoop.
Myth 2: Its only useful for personalization
Companies like Facebook use Big Data to decide which newsfeed ads to display to users, but this technology isnt only about serving just the right thing in context. In fact, analytics are most useful for answering specific questions about a puzzling customer behavior that could be symptomatic of an underlying problem in the business. Thats especially true in crowded marketplaces where product owners must analyze everything in the quest to reduce the end users time to value.
Myth 3: We need a big data team
From an organizational design perspective, your companys data science practitioners should indeed share processes, definitions and culture. But the individuals themselves should sit with the teams theyre interacting with, whether theyre business teams, product teams or R&D. Embedding data scientists with cross-functional teams lets them absorb the working knowledge of the domain experts that most need their input, which results in anticipatory collaboration instead of trial and error.
Myth 4: We need powerful technology to get answers
One of the most promising technologies in use today is machine learning, which can reproduce human judgment in sorting functions sifting through a million emails for criminal evidence during an investigation. But businesses begin implementing machine learning workflows before theyve figured out the right business questions to ask. Big Data isnt useful if it is generating answers to things nobody cares about. An easy way to ensure your team uses analytics intelligently is to begin with simple statistics. Someone with business savvy can load a data set into Microsoft Excel, run a basic regression analysis and reveal causes or predictors sitting right under your teams noses, giving your data scientists a better idea of where to dig deeper down the road.