Commentary
When The Data Tells All You Need To Know
- by Laurie Sullivan , Staff Writer @lauriesullivan, January 9, 2013
I love data like I love country music. I admit it. While country music makes you feel good, data answers so many questions. After trying for months to answer the question of why I kept seeing an abundance of country-music-related television shows, I came across a number from Nielson SoundScan that answers the question.
It turns out the country genre demonstrated the greatest gains in 2012, with sales up 4.1% compared with the prior year. Not even rock, which continues to be the biggest genre, could surpass country with 1.4% gains. Country digital album sales also rose 38% in 2012 vs. 2011, the largest increase of any genre in the digital album format.
Perhaps that's why Jason Aldean will perform at the People's Choice Awards Wednesday night. Consumers show intent and brands need to deliver. How does a marketer follow through without the numbers to back up the claim?
comScore cofounder Gian Fulgoni's post gives us the top 10 "burning issues in digital," with big data listed as No. 1. He tells us 120 million people in the U.S. now own smartphones, up 30 million in just the past year; that for $600 you can buy a disk that can store all of the world’s music; and 92% of the world’s data was created in just the past two years. What can you do with that knowledge?
advertisement
advertisement
Forrester Research Analyst Mike Gualtieri believes big data, modeling tools and algorithms will become key differentiators in 2013, and the market for such services will become "highly competitive," with an overabundance of new entrants during the next three years.
It's pretty darn simple. Predictive analytics uses algorithms to find patterns in data that might predict similar outcomes in the future, according to Gualtieri. He explains that predictive analytics allows telecom carriers to predict the customers most likely to switch carriers based on number of calls made, minutes used, number of texts sent, average bill amount, and hundreds of other variables.
The key is to continually rerun the data, because consumer intent changes will prove successful when brands set business goals, understand data from a variety of sources, create a predictive model, and more. And while the Forrester report--The Forrester Wave: Big Data Predictive Analytics Solutions, Q1 2013--details possible leaders in the space, such as SAS, IBM, and SAP and Oracle, it also provides a roadmap for companies looking to capitalize on data sources.
Laurie Sullivan is a writer and editor for MediaPost. You can reach Laurie at lauriesullivan@gmail.com.
In the end, my guess is big data will end up having been a massive distraction. From what I've seen so far, the vast majority of times there aren't massive leaps in effectiveness to be gained - merely some small increments.
Classic problem that does not exist - Big Data.
Analogy: You don't need to understand wind patterns, velocities and directions by studying every molecule of air - you build a model to sample attributes at sufficient granularity to be predictive.
The answer lies in using the right statistical approaches and filtering filtering filterin out noise.
Would also add that capturing data into storage areas is stupid. You need to extract metadata from the data in "motion" for in memory analytics and then store a representation of the results in your databases (data at "rest").
Data from smart energy meters are a classic example of the above situation.
I think we need a new name. The big data processed by predictive analytical solutions such as those provided by SAS, IBM, SAP, and Oracle is in fact big and those vendors really can extract value from it. However… there is a class of data that is even bigger. This bigger data shouldn’t be stored and can’t be queried. As Andre Szykier correctly points out, this data is so big that it has to be processed in motion so that at least part of the algorithms move into the detection and collection processes. Different data structures and different processing techniques are required and commercially available predictive solutions do not apply. When collected in the pursuit of online advertising results, the data is not only bigger, but also sparser. Sparser refers to the fact that there are so few responses compared to the possible sites and consumers that sampling the data down a size eliminates all of the value. It is possible to “produce massive leaps in effectiveness” from bigger data, a select few ad tech providers do it every day, but it takes purpose built systems that implement their algorithms across the entire architecture from monitoring and collection through storage and analysis to scoring in real-time. Those who employ general purpose predictive solutions or apply algorithms in an analytical step, will get “small increments of improvement” at best as Doug Garnett correctly points out
Data alone will never tell you all you nee to know about anything