Click here to load reader
Ask Measure Learn
Jun 19, 2015
Making sense of your data doesn't require mountains of data; it requires a systematic approach that leads to actionable insights. But how to get there? This keynote (given at "Predictive Analytics and Business Insights") shows how to extract significant business value from big data with Ask-Measure-Learn, a system that helps you ask the right questions, measure the right data, and then learn from the results. Using this system can help you learn to:
* Focus on business-related questions
* Find measures that have high causation, a low error rate and a low cost
* Create actionable insights by starting with predictions, benchmarks or recommendations
* Focus on business-related questions
* Find measures that have high causation, a low error rate and a low cost
* Create actionable insights by starting with predictions, benchmarks or recommendations
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Ask Measure Learn Lutz Finger
Last update: March 3, 2014
The philosophy of the day is . data-ism
—DAVID BROOKS (@NYTDAVIDBROOKS)
Who ? • Lutz Finger • Director Data Analytics at LinkedIn • Author of Ask Measure Learn • Co-Founder Fisheye Analytics
(sold to WPP) @LutzFinger LutzFinger.com linkedin.com/in/lutzfinger
am I
Who ? are you
Rahul Todkar Most Popular
Mario Vinasco Earliest Adopter
Bruno Aziza Most Connected
Ron Goncalves, MBA, PMP,
Most Endorsed
Actionable Insights?
Intellectually enriching – at most
We focus too much on technology
ENGINEERS DATA SCIENTISTS
Google Search on the term “Big Data”
ASK the right Questions.
MEASURE the right data – even if it is not Big data.
Take Actions and LEARN from them.
?
Ask the right
Question
ASK
THE HARDEST PART
ASK
Let’s do “Social Media”
ASK
ASK
Source: IBM Institute for Business Value.
Please find an “INFLUENCER”
• Opinion leaders (Katz 1955) • Influentials (Merton 1968) • Law of the Few (Gladwell 2000)
ASK
Source: ‘Ask Measure Learn’ by O’Reilly Media
A few person decide what we do…
ASK
Source: ‘Ask Measure Learn’ by O’Reilly Media
REALLY?
ASK
Source: ‘Ask Measure Learn’ by O’Reilly Media
Dear Marketers,
There is no influencer. It’s a myth.
ASK
Reality 50% is Homophily
Source: Kevin Lewisa, Marco Gonzaleza and Jason Kaufman (2012): PNAS Vol 109, no 1
• 4 years • 1001 Students on Facebook • traditional Self-reported Data • How did taste Spread
Influence is often overestimated. It needs:
§ Reach § Readiness
§ Topic Dependence
ASK
Aja Dior M.?
AP News?
ASK
Aja Dior M. omgg, my aunt tiffany who work for whitney houston just found whitney houston dead in the tub. such ashamed & sad :(
45 min
Aja Dior M.?
AP News?
ASK
I want to create “REACH”
… in order to “SELL”
ASK
Measure the right Data
MEA
SURE
The world is full of WRONG metrics
• Click Rates – Media • Low CPO advertisers • Etc..
MEA
SURE
30
What is RIGHT?
Mea
sure
M
EASU
RE
Source: WIkipedia
even if it is not Big Data
MEA
SURE
277+ Million Members = a lot of Data
Data Volume
Ana
lytic
al C
halle
nge
&
Com
ple
xity
Identity Data
Social Data
Behavioral Data
MEA
SURE
More Data More Insights does not equal
MEA
SURE
Web API
Data ELT and Aggregation
Petabyte Gigabyte Bit Terabyte 1 2 3 4
We want small data….
Analytics Portal
Reach or
No Reach
MEA
SURE
Calculating Reach via Network Data
1977: Linton C. Freeman, “Centrality based on Betweenness .”
MEA
SURE
“REACH” creates awareness
“SELL” needs purchase intend
MEA
SURE
Correlations are important
MEA
SURE
What is your behavior?
MEA
SURE
Source: ‘Ask Measure Learn’ by O’Reilly Media
The issue with the Correlation / Causation
MEA
SURE
Sometimes data is not easy to get.
MEA
SURE
Social Behavior is ‘unstructured’
MEA
SURE
Source: ‘Ask Measure Learn’ by O’Reilly Media
It is way easier to work with ‘structured data’
New York Weather in April 2013
MEA
SURE
Source: ‘Ask Measure Learn’ by O’Reilly Media
43
ME
AS
UR
E
MEA
SURE
Source: Jeffrey Breen
44
What is RIGHT?
MEA
SURE
Source: ‘Ask Measure Learn’ by O’Reilly Media
Right and learn from them.
Actions
LEA
RN
Information vs. Action…
0 Connections
59%
1 Connection
28%
2 Connections
4%
4 Connections
3%
LEA
RN
47
Information vs. Action…
0 Connections
59%
1 Connection
28%
2 Connections
4%
4 Connections
3%
LEA
RN
Three Types of actionable Insights
Benchmark Predictions
Recommendations &
Filter
LEA
RN
BENCHMARK
LEA
RN
50
Competitive Benchmark
LEA
RN
Source: ‘Ask Measure Learn’ by O’Reilly Media
RECOMMEND & FILTER
LEA
RN
LinkedIn Recommendation Products
People You May Know
Groups You May Like
Ads You Be Interested in
Companies you May Want to Follow
Puls
Similar Profiles
LEA
RN
Recommendations Impact
More than 50% of
group joins
More than 50% of
connections
More than 50% job
applications
LEA
RN
Filter
LEA
RN
Source: ‘Ask Measure Learn’ by O’Reilly Media
PREDICTIONS
LEA
RN
Predicting the OSCAR 2013
LEA
RN
Predicting the OSCAR
LEA
RN
Predicting the OSCAR
LEA
RN
Source: Farsite
Possible other Features: CROWDSOURCED: • Box Office Results • Movie Goer Reviews • Critics OTHER “Hard Facts” • Genre • Payment of Actors • Etc.
Predicting Box Office
LEA
RN
Source: ‘Ask Measure Learn’ by O’Reilly Media
Predictions are not easy, especially if they are about the Future
LEA
RN
Propensity to Email Marketing
Emily
Score = 0.9
Steve Score = 0.9
Behavioral Score = 0.5
Identity Score = 0.3
Social Score = 0.1
Behavioral Score = 0.2
Identity Score = 0.3
Social Score = 0.4
... With premium account, you get more search results and access to …
... Do you know 5 of you connections have started to use premium account …
LEA
RN
Which Model to use
LEA
RN
Source: ‘Ask Measure Learn’ by O’Reilly Media
Which Model to use
LEA
RN
Google’s Prediction API
Many more examples…
Benchmark Recommendation & Filter Predict LEA
RN
Data Wrangling & Data Science is getting Easier
LEA
RN
Update Training
LEA
RN
Summary
ASK the right Questions.
Forget Big Data but
MEASURE the right data.
Take Actions and LEARN from them.
?
We are hiring!
Related Documents
