Big Data refers to the large amounts of data being created from various sources such as mobile devices and social media. This data presents opportunities for personalization, prediction, and prevention through analyzing trends and correlations. To get started with Big Data, companies should focus on integrating their existing data from various sources and improving data quality before applying more advanced analytical techniques.
3. Housekeeping
Please... keep your phones ON and go to:
pollev.com/jmck
Feel free to put them on ‘silent’ though :-)
pollev.com/jmck
From
any
browser
3Wednesday, 20 August 14
4. Agenda
• What is Big Data?
‣ Where did it come from? Why now?
• What opportunities does it present?
‣ Personalisation, Prediction, Prevention
• How do I get started?
• NOTE: This is an opinion piece (It’s not a science!)
4Wednesday, 20 August 14
9. Why Big Data?
• 7 billion people access 6 billion mobile devices
• Last year we...
‣ Sent 11 billion texts
‣ Watched 2.8 billionYouTube videos
‣ Performed 5 billion Google searches
• The world’s data doubles every 2.1 years
9Wednesday, 20 August 14
10. Where is it coming from?
• Increased device accessibility
• New storage paradigms
• New transaction types
• Growth in social media
• Increase in use of rich media
• More conversations
10Wednesday, 20 August 14
12. The Internet Of Things
• Internet-enabled everything
• Objects predict their own failure
‣ ... and wirelessly notify their manufacturers
‣ ... who automatically pre-order parts
• Objects upgrade themselves
‣ ... such as this Mac
• Objects communicate with one another
‣ Energy companies will control demand
12Wednesday, 20 August 14
14. Why Invent ‘Big Data’?
• Big Data is about more than just ‘lots of data’
‣ ... although that’s part of it
• Big Data typically characterised by ‘3Vs’:
Volume
Variety
Velocity
14Wednesday, 20 August 14
15. Volume
• Typically measured in Petabytes
‣ A gigabyte is 7 minutes of HD video
‣ A terabyte is 120 hours of HD video (1024 Gb)
‣ A petabyte is 14 years of HD video (1024 Tb)
• Accelerating rate of growth - driven largely by mobile devices
• Prices dropping dramatically
15Wednesday, 20 August 14
16. Storage is Cheap
• Storage costs are reducing
exponentially
• Data expands to fill the space
available
• Heading fast towards the online
‘Personal Petabyte’ 0.00001
0.00010
0.00100
0.01000
0.10000
1.00000
10.00000
100.00000
1,000.00000
10,000.00000
1980 1989 1997 2006 2014
Source: PC magazine, Byte magazine, newegg.com
Storage Costs ($US/Mb)
16Wednesday, 20 August 14
17. Variety
• Traditional databases are designed for well-structured data
• Making sense of free-form text?
• Extracting information from audio?
• Searching video?
• New relationship structures between data
‣ Increasing use of network modelling
17Wednesday, 20 August 14
18. Network Modelling
• Sentiment is viral!
• Uncovers relationships of varying types
and strengths
• What are the distinct groups within your
customer networks?
• Who are the most and least connected?
• Who are the ‘influencing nodes‘ in your
customer networks?
18Wednesday, 20 August 14
20. Why Look At Networks?
• Unhappy customers vent frustrations on social networks
• Those using Twitter are already disproportionately upset
‣ Compared to those raising traditional complaints
• Twitter complaint response: 3 minutes ➔ 70 minutes
• Email complains: 24 hours (30%) ➔ Never (70%)
‣ Almost ¾ of organisations are ignoring their customers!
20Wednesday, 20 August 14
21. Viral Complaints
• The Dave Carroll band were flying with United Airlines whose
handlers damaged his guitars
• Complaints were met with rudeness, avoidance and red tape
• HisYouTube response video received 13 million hits
• Negative sentiment flooded social networks
• United Airlines’ stock dropped 10% ($180 million)
21Wednesday, 20 August 14
22. Velocity
• Driven by proliferation of mobile devices
• Twitter processes over 34,000 tweets every 60 seconds
• Amazon process approximately 20 million transactions a day
• The SKA, due for completion in 2024, will generate...
‣ 1,376 petabytes per day
‣ Twice the current daily global internet traffic!
22Wednesday, 20 August 14
23. What’s Big Data?
• ‘Traditional’ data processing technology isn’t designed for Big Data
‣ Ask Facebook, Google,Twitter, eBay, Amazon,Walmart, ...
• Big Data could be thought of as an organisational toolkit:
‣ Application of new technologies to handle the 3V’s
‣ Application of advanced statistical tools to our data
‣ Adaptation of business processes to leverage new insight
23Wednesday, 20 August 14
27. Predicting Politics
• Nate Silver
• Big Data Scientist who started by
predicting baseball results
• Famous for predicting 2008 US
election results with 98% accuracy
• Did it again 2012 with 100%
accuracy (predicted Obama 91%)
27Wednesday, 20 August 14
28. Predicting Crime
• “PredPol” predictive policing
initiative
• Los Angeles Police Department
and the University of California
• Software predicts where crime
will occur within a given area
• Based on analysis of 13 million
crimes over the last 80 years
28Wednesday, 20 August 14
29. Predicting Crime
• Mathematical model originally
determined the location of
earthquake aftershocks
• Crime prediction model is
updated with new crime data in
real time to improve accuracy
• Result: 12% decrease in property
crime, 28% decrease in burglary
29Wednesday, 20 August 14
30. NSW Police 3rd Eye Cameras
• Sydney police getting vest-
mounted cameras
• The Wolfcom units record
- 6 hours of HD video
- 20K 12 megapixel images
- 500 hrs voice recording
- All GPS tagged
• How is this used?
30Wednesday, 20 August 14
31. Target
• Minneapolis father furious at ‘offensive’ marketing to his daughter
• ... due to Andrew Pole, Big Data specialist at Target
• Andrew identified about 25 products that, together, allowed him to
assign each online user a ‘pregnancy prediction’ score
• ... which can also estimate the due date to within a few days!
• Target uses this to send coupons timed to very specific stages of
pregnancy
31Wednesday, 20 August 14
33. Holistic Data
• Traditional approaches used data sampling due to data volumes
‣ Take every n’th record
‣ Take selected records (e.g. geographical or other segments)
• Sampling is often biased
‣ Statistical aberrations
‣ Simpson’s paradox
33Wednesday, 20 August 14
34. Unstructured Data
• Incorporate unstructured data into your analysis
‣ Twitter, Facebook, Social Media
‣ Emails, Contact notes
‣ Audio, Pictures,Videos
‣ Networks
• Distill these and use them to feed analytic models
34Wednesday, 20 August 14
35. Correlation over Causation
• Traditional analysis involves testing hypotheses against our data
‣ Requiring a hypothesis
‣ Root cause analysis based on guessing reasons for behaviour
• Holistic data opens the door to a new approach
‣ Focus on influencing factors rather than possible causes
‣ Root cause analysis based on statistical probability
35Wednesday, 20 August 14
40. Big Data for Complaints
• Anticipate complaints
‣ Based on statistical probability and our customer insight
• Identify the root cause of complaints
‣ Link complaints to business processes and organisational change
• Proactively engage customers in high quality conversations
‣ So poor conversations don’t escalate into complaints
40Wednesday, 20 August 14
41. The Opportunities
• Derive insight from customer behaviour
• Analytic probabilities rather than traditional signals
‣ Correlation over causation
• Augment our data with 3rd party intelligence
• Derive insight from non-traditional (unstructured) sources
41Wednesday, 20 August 14
42. How Do I Get Started
With Big Data?
42Wednesday, 20 August 14
43. Statistically Probable Starting Point
• Big Data is not a panacea!
• Big Data will not fix your data quality issues
‣ Customer insight requires a single customer!
• Start by assessing your current information architecture
• Data Integration
43Wednesday, 20 August 14
47. Getting Started
• Focus on...
‣ Data Integration
‣ Data Quality
‣ Master Data Management
• Big Data can help these initiatives
‣ ...but you need to reach a minimum threshold before you start
47Wednesday, 20 August 14
49. What can Big Data do for me?
• You don’t need a SKA or 1.23 billion users (like Facebook) to
benefit from the approaches adopted by the Big Data organisations
• The Big Data toolkit incorporates...
• Technology adoption
• Statistical modelling
• Business change
49Wednesday, 20 August 14
50. What can Big Data do for me?
• Understand your customers
• Customer segmentation to a segment of one - The Customer
• Anticipate their needs
• ... and hence their behaviour
• Drive high quality conversations with them
• Based on your understanding of them
50Wednesday, 20 August 14
51. Big Data Technologies
• Business intelligence
• Visualisation
• Infrastructure
• Agile methodologies
• New data storage
architectures
• Parallel processing
• Machine learning
• Statistical modelling
51Wednesday, 20 August 14
52. Approaches Opportunities
Big Data Summary
Characteristics
• Velocity • Incorporating
unstructured data
into your analysis
• Holistic data rather
than sampling
• Correlation rather
than causation
• Complaints root
cause analysis
• Volume
• Improve the quality
of conversations
• Anticipate
behaviour through
deep understanding
• Variety
52Wednesday, 20 August 14
53. Privacy - Social Media
• Legislation will always trail technology
• Social Media sites most frequently have “a worldwide, non-
exclusive, royalty-free license, with the right to sublicense”
• You’ve never paid Facebook or Twitter a cent
• They can ‘monetize’ both your content and your metadata
• Most legislation centres around self-policing and opt-out
53Wednesday, 20 August 14
54. Privacy - Individuals’ Rights
• Staples (US) operate a punitive pricing model
‣ Your IP address tells them if you live in an
expensive neighbourhood
• OfficeMax addressed marketing to
”Mike Seay, Daughter Killed in Car Crash”
• Washington D.C. Police office convicted after
looking up licence plates of vehicles near a
gay bar and blackmailing the vehicle’s owners
54Wednesday, 20 August 14
55. Privacy - I can buy your...
• Full name, spouse, children, ex-partners, co-habitees, current
address, previous addresses, ownership status, purchase date and
price, outstanding mortgage debt
• Job type, income band, credit score, credit and store cards, spending
habits, charitable contributions, family events (births, deaths) and
likely political affiliation
• Ethnicity, primary language, and (in the US) health information!
‣ Cancer, diabetes and clinical depression lists with credit score
55Wednesday, 20 August 14