How to Get Your Brand Elected: Big Data Presentation at NRF by David Selinger
1. How Relevance Can Get
Your Brand Elected
DAVID SELINGER
daveselinger@richrelevance.com
www.richrelevance.com
RAYID GHANI
rayid@uchicago.edu
www.rayidghani.com
3. Our Platform
1. The History of Data
2. How Data Make Money, Win
Elections and Other Such Nonsense
3. Demystifying the Mystical Creature,
Big Data
4. Next Steps: How to Take the Next
Steps
#richrelevance
16. 1
2
Prediction & Interruption
3
Key Campaign Lessons
Analytics
Optimization
Q: Who are the best people to persuade
and influence?
A: The people who are just like you
#richrelevance
24. Big Data: The 3 V’s
Velocity
#richrelevance
Volume
Variety
25. Surprise!
HOW WE GOT ELECTED WITH DATA:
Lots of good ol’ data integration with sprinkles of Hadoop
dust and other” big data” technologies when appropriate.
#richrelevance
30. How do we apply these to our businesses?
• Getting Started: Learn Quickly
• Getting Integrated: Bite-sized Chunks
• Measure yourself: The right metrics
• Experiment, Experiment, Experiment.
• Scale: Rinse and Repeat or “Big Project”
#richrelevance
31. Some Initial Big Data Retail Projects
•
•
•
•
SUCCESSFUL
Multi-channel marketing
attribution
Social network
demographics for content
targeting
Multi-channel customer
profile creation for
analytics, targeting,
personalization
Datamart creation
#richrelevance
•
•
•
•
UNSUCCESSFUL
Data warehouse
replacement
“Big Data Experiment”
Online
analytics/segmentation
Board of Directors “Check
the Box” Project: “Yes,
ma’am we’ve got Big
Data”
33. Thank you!
ACCESS THE HANDOUT:
richrelevance.com/nrf
CONTINUE THE CONVERSATION:
DAVID SELINGER
To track the progress
of Big Data for
retailers, subscribe to
our regular newsletter
“Big Data for Retailers”
retail-bigdata@
richrelevance.com
#richrelevance
daveselinger@richrelevance.com
www.richrelevance.com
RAYID GHANI
rayid@uchicago.edu
www.rayidghani.com
Editor's Notes
AnalyticsInsightsReportingPrediction & InterruptionFraud detectionProduct recommendationsOptimizationTweakingHill-climbing
(no version 2.0, no better quarterly earnings), & there’s no “PERFECT” but keep going (Hill Climbing!)
(but might be like YOUR org! (aligning/motivating the behemoth & personalizing at vast scales)
http://en.wikipedia.org/wiki/Paul_the_Octopus
AnalyticsInsightsReportingPrediction & InterruptionFraud detectionProduct recommendationsOptimizationTweakingHill-climbing
They subsidize with advertising.
Product/Infrastructure: 5 projects/yearIT/Finance: IT budget as % of salesTalent: Recruiting against FB, Google, AmazonPace of Innovation: Same 5 projects…Extensive economies of scale
Velocity: Data are moving faster—real-time means real-value.Volume: 100’s of PB’s of data vs TB’s of data—e.g., Walmart’s 1995Variety: Social data is messy, rich, and valuable, but how the heck do we work with it?!??!
How we got elected with data:We used mostly basic data storage tech—and Hadoop only when appropriate.
SELLY Data can be big and dumb--that is unwieldy for the task at hand ... or pursued just because it CAN be, with no thought for value. Especially when 10X the data doesn't mean 10X the insight (i.e. the law of diminishing returns), sometimes small data is just fine. What your data has to be 24/7 is SMART. (Taken from Forbes “In Defense of Big Data” article) Consider the images on this screen to illustrate. Is there more information in the high-resolution picture at the left or the middle? Absolutely! But although that image depicts a 10X bigger data set, it’s not conveying 10X the insight. If I had to speculate, I’d say that informational content is proportional to the logarithm of the quantity of data – a fancy way of reiterating the law of diminishing returns.My third and final point is that one should focus on the diversity of data available. Don’t give me 10X or 100X or 1000X finer detail. Instead, I derive far more insight by introducing data of a different type even if it’s small data. To illustrate: low resolution with color tells a richer story than the high resolution black-and-white.But data can’t just be big. It has to be timely. “We’re in a world now where you can’t wait a whole day to get your data,” says Julie Kim, VP of product at the furniture etailer One Kings Lane. President of Retail Prophet,Doug Mack, agrees that feedback speed is vital: “Big data is not enough. It has to be real-time data that we can act on during the course of the day.” And to be actionable, data must be personal.
Learning, Working on it, Integrated into work, Dependent upon it, DNA & Culture