ray poynter big data and advanced analytics
TRANSCRIPT
Ray Poynter, The Future Place – JMRX Lectures 2015
Big Data and Advanced Analy0cs
Ray Poynter The Future Place
JMRX – Tokyo – May 12, 2015
Ray Poynter, The Future Place – JMRX Lectures 2015
Agenda
1. What is big data?
2. The strengths of big data 3. When big data misses the mark
4. The challenges of big data 5. Linking big data, analyEcs and market
research
6. ImplicaEons for Japan
Ray Poynter, The Future Place – JMRX Lectures 2015
How big is big data? BIG! • If it fits in Excel, it is not really big data • If it fits in SPSS, it is not really big data • If you have a profile of 10 million telco customers, it is not really big data
• If you have second-‐by-‐second locaEon and usage data for 10 million telco customers, that is big data
• If you need to use Hadoop, it is big data
Ray Poynter, The Future Place – JMRX Lectures 2015
What is ?
InstrucEon
Results
Ray Poynter, The Future Place – JMRX Lectures 2015
IBM’s four Vs
Ray Poynter, The Future Place – JMRX Lectures 2015
What does Google Know?
And shares with you
Ray Poynter, The Future Place – JMRX Lectures 2015
Google Ads http://www.google.com/settings/ads/
Ray Poynter, The Future Place – JMRX Lectures 2015
Loca0on History https://maps.google.com/locationhistory
7 days of travel
Ray Poynter, The Future Place – JMRX Lectures 2015
Google Search History https://www.google.com/history/
Ray Poynter, The Future Place – JMRX Lectures 2015
Google Other
4. Google monthly security and privacy report – lisEng all the services you use hVps://www.google.com/seYngs/dashboard
5. Apps and extensions that have access to your Google data hVps://security.google.com/seYngs/security/permissions
6. Export all your Google data hVps://www.google.com/takeout
7. History of all your YouTube searches hVps://www.youtube.com/feed/history/search_history
Ray Poynter, The Future Place – JMRX Lectures 2015
The Signal and the Noise – Nate Silver
Ray Poynter, The Future Place – JMRX Lectures 2015
Weather forecas0ng
PMSL – Pressure at Mean Sea Level
Ray Poynter, The Future Place – JMRX Lectures 2015
Weather forecas0ng benchmarks
1. Same as today
2. Same as average of last few years -‐ climate
What benchmark are you going to use for your predictive analytics?
Ray Poynter, The Future Place – JMRX Lectures 2015
Google Flu Trends
http://www.google.org/flutrends/ December 2014
Ray Poynter, The Future Place – JMRX Lectures 2015
Google Flu Trends
http://www.google.org/flutrends/ May 2015
Ray Poynter, The Future Place – JMRX Lectures 2015
Google Flu Trends -‐ Japan
Ray Poynter, The Future Place – JMRX Lectures 2015
Target
Ray Poynter, The Future Place – JMRX Lectures 2015
Ray Poynter, The Future Place – JMRX Lectures 2015
BPP and USA Prices Index
Ray Poynter, The Future Place – JMRX Lectures 2015
Tesco • Loyalty card data
• Real-‐Eme monitoring of refrigerators across 120 stores in UK and Ireland to save 20million Euros a year – 70 million data points
• Real-‐Eme monitoring of lighEng and heaEng, across 120 stores – Management dashboard shows by 7am which stores are not at the right temperature
• Buying SocialmaEcs – programmaEc and re-‐targeEng ads
Ray Poynter, The Future Place – JMRX Lectures 2015
INTERESTING BIG DATA PROJECTS
Ray Poynter, The Future Place – JMRX Lectures 2015
Ebola and mobile phones
Flowminder using data from telcos to map populaEon mobility – but level of granularity is quite large – journeys over 20KM
Ray Poynter, The Future Place – JMRX Lectures 2015
The smartphone Accelerometer
Temperature
Gravity
Gyroscope
Light
Air pressure
Proximity
Humidity
GPS
Call acEvity
App acEvity
Internet usage
WiFi
Bluetooth
Cameras
Near Field Comms
GSM/CDMA
Ray Poynter, The Future Place – JMRX Lectures 2015
Aberdeen tracks ‘hundreds of thousands’
Ray Poynter, The Future Place – JMRX Lectures 2015
Kingsgate Shopping Mall Huddersfield, Yorkshire (UK)
Ray Poynter, The Future Place – JMRX Lectures 2015
Mount Sinai Hospital
Mount Sinai are using Big Data approaches to uElise data to make personalised health and treatment predicEons. Project led by Jeff Hammerbacher, a 30-‐year-‐old known for being Facebook’s first data scienEst.
Ray Poynter, The Future Place – JMRX Lectures 2015
Big Data Failures
Ray Poynter, The Future Place – JMRX Lectures 2015
The Signal and the Noise – Nate Silver
Ray Poynter, The Future Place – JMRX Lectures 2015
Earthquakes
Given the abruptly self-organizing nature of earthquakes, it is extremely unlikely that precursors can attain such levels of accuracy. I therefore conclude that prediction of major earthquakes is, in any practical sense, impossible. Russ Evans, 1998
Ray Poynter, The Future Place – JMRX Lectures 2015
Big Data thinks I am a pregnant dude!
http://www.carolroth.com/blog/google-dude-target-pregnant/
Ray Poynter, The Future Place – JMRX Lectures 2015
Ray Poynter, The Future Place – JMRX Lectures 2015
Google Flu – the failures
Ray Poynter, The Future Place – JMRX Lectures 2015
Flu Near You – a crowdsourced approach
Ray Poynter, The Future Place – JMRX Lectures 2015
The Samaritans App
Ray Poynter, The Future Place – JMRX Lectures 2015
ScrapeGate
Ray Poynter, The Future Place – JMRX Lectures 2015
BIG DATA CHALLENGES
Source: Gizmodo
Correla0on Annual Chocolate Consump0on & Nobel Prizes per 10 Million of Popula0on
New England Journal of Medicine.
Ray Poynter, The Future Place – JMRX Lectures 2015
Correla0on and Causa0on
1. CorrelaEon predicts the past – Which is someEmes enough – Especially when the past repeats itself
2. CausaEon is needed to predict new futures – But causaEon is hard to establish in the real world
3. Experiments are key – Market research can help
Ray Poynter, The Future Place – JMRX Lectures 2015
Perverse incen0ves
• Colonial Hanoi – bounEes for rats tails • Duplessis Orphans – Canada 1945-‐60, Orphans=70 cents,
mentally ill $2.25, 20K children confined • Paying for acEon (e.g. doctors and firemen) decreases
prevenEon work • Facebook likes correlate with success – unless they are set as
a target, creaEng a black market in them • TwiVer menEons correlate with success – unless they are
benchmarked, creaEng value in them • Big Data benchmarks change behaviour to improve scores,
rather than underlying performance
Ray Poynter, The Future Place – JMRX Lectures 2015
Where are the data scien0sts going to come from?
Google Trends
Ray Poynter, The Future Place – JMRX Lectures 2015
MARKET RESEARCH AND ANALYTICS
Ray Poynter, The Future Place – JMRX Lectures 2015
A\ribu0on Modelling
Ray Poynter, The Future Place – JMRX Lectures 2015
Knowing where to dig and digging Lucien Bowater, Director Strategy and Insight at BSkyB – UK media company MRS Conference, UK, 2013
Issue MR Digs Big Data AcEon CEO
PaVern Big Data Digs MR AcEon CEO
Ray Poynter, The Future Place – JMRX Lectures 2015
T hVps://www.ted.com/talks/ben_wellington_how_we_found_the_worst_place_to_park_in_new_york_city_using_big_data
Ray Poynter, The Future Place – JMRX Lectures 2015
Ray Poynter, The Future Place – JMRX Lectures 2015
Where to dig and digging
1. When is rush hour? MR asks the quesEon, Big Data digs
2. Why is this fire hydrant generaEng so much money in parking Eckets? Big Data asks the quesEons, but it was Qual that answered it.
Ray Poynter, The Future Place – JMRX Lectures 2015
Thank You!
Questions?
Ray Poynter, The Future Place – JMRX Lectures 2015
IMPLICATIONS FOR JAPAN?
1. What new skills do we need? 2. What new people do we need? 3. What new tools do we need? 4. How should MR work with Big Data
and Analy0cs?