Erin Shellman, PhD 
@erinshellman 
! 
Marketing Analytics 
March 18, 2014 
real talk. 
real-time
Hello.
real talk. 
real-time 
1. real-time - what? 
2. real talk - look, baby. 
3. Nordstrom - real good.
Big data was last year, 
now we want fast data.
“…Real-Time Analytics is 
enabling retailers and 
brands to analyze the 
right data in real-time to 
generate solutions that 
can be implemented 
across an organization1.”
What is a real-time 
system? 
A real-time system is one that must 
satisfy bounded response-time 
constraints or risk severe 
consequences, including failure2.
What is real-time 
analytics?
In general, real-time analytics can 
be defined as the use of, or the 
capacity to use, available 
enterprise data and resources 
when needed3.
“It’s about moving with greater 
speed toward previously unknown 
questions, defining new insights, 
and reducing the time between 
when an event happens 
somewhere in the world and 
someone responds or reacts to 
that event4.” 
— Justin Erickson, Cloudera
“Typically when we’re talking about 
real-time or near real-time 
systems, what we mean is 
architectures that allow you to 
respond to data as you receive it 
without necessarily persisting it to 
a database first.” 
— John Akred, Accenture
Real-time analytics 
in the wild.
Williams Sonoma 
integrates its customer 
databases with external 
demographics to 
created targeted 
emails. 
e.g.
e.g. 
Macy’s personalizes 
their catalogs to the 
point where they could 
deliver 500,000 
unique versions of a 
single direct mail 
catalog.
e.g. 
Koupon Media built a 
real-time dashboard 
enabling them to 
identify key attributes 
of users who 
participated in their 
marketing campaigns.
Are any of these use-cases 
meaningfully real-time?
Nope!
Nope!
Nope! 
Real Talk!
1. If you can’t act on it in real-time, 
it’s not really real-time.
1. If you can’t act on it in real-time, 
it’s not really real-time.
1. If you can’t act on it in real-time, 
it’s not really real-time. 
Real Talk.
“If you have people in the 
loop, it’s not real-time.” 
— Joe Hellerstein, UC Berkeley
2. Real-time is not the 
same as all-in-one-place.
2. Real-time is not the 
same as all-in-one-place.
2. Real-time is not the 
same as all-in-one-place. 
Real Talk.
3. Real-time systems fail 
when messages are late.
3. Real-time systems fail 
when messages are late.
3. Real-time systems fail 
when messages are late. 
Real Talk.
IBM helped German retailer 
METRO boost customer 
satisfaction by 18% with 
intelligent dressing rooms 
that detect the shopper's 
current apparel selection and 
make appropriate 
recommendations for 
accessories. e.g.
E-mail offers from Microsoft 
to engage with Bing are 
customized at the time of 
opening. In 200 milliseconds, 
an offer is assembled based 
on the most recent customer 
responses and the available 
real-time information about 
the current customer – 
leading to up to a 70% 
e.g. increase in conversion rates.
What’s the difference? 
• Real-time systems are largely automated, end-to-end. 
• Building real-time systems requires deep insight, and 
not so much the other way around. 
• Real-time analytics is not the push-button solution it’s 
pitched as.
Recommendo 1.0
Recommendo 2.0 
personalized recommendations in real-time!
Batch Real-time 
Purchase history 
Demographics 
Category affinity 
Brand affinity 
Collaborative 
Filtering 
Items 
Users 4 … 
… 0 
Clicks 
Logistic 
Regression 
Recs Probability 
Item 1 0.15 
Item 2 0.01
Target 
Product 
Product-to-product Personalized Batch Session-driven 
Reorder Reorder 
Best sellers Prefers blazers Clicks red & orange 
3 recs per page 
1 
2 
3 
4
What’s an enterprise to do? 
• Be disciplined. 
• Insight before automation. 
• Fully invest in data technology!
Being data-driven is a hot phrase: but what 
does it mean in practice? It’s not about 
finding analytical fairy dust that you can 
sprinkle to make everything better, but a 
deep-rooted approach to operations, 
innovation and competition. Creating an 
organization designed for data-informed 
decision—the experimental enterprise—is 
the key to competing in a world that’s 
being rapidly digitized. 
— Edd Dumbill, Forbes
Thanks pals! 
! 
Sara Hogenson - Designer, Nordstrom Innovation Lab 
Scott Jones, Prashanth Nair, Abid Saifee 
The Advanced Analytics, Personalization, and Mobile Teams 
Michael Draugelis - Lockheed 
R. Kelly - Real talk
Learn more 
1. PSFK Future of Retail trend report. 
2. Laplante, Phillip A. “Real-time systems design and analysis.” 
3. SAP: http://blogs.sap.com/innovation/analytics/real-time-data-analytics-09636 
4. Barlow, Mike. “Real-time big data analytics: emerging architecture.” 
5. Infochimps: http://www.infochimps.com/resources/koupon-media-case-study/ 
6. “Real-Time Retail: Rapid Response to Consumer Demand” from the International Institute for 
Analytics’ Retail Analytics Research Council. 
7. Dumbill, Edd. Forbes. http://www.forbes.com/sites/edddumbill/2014/02/05/the-experimental-enterprise/? 
imm_mid=0b76e4&cmp=em-strata-na-na-newsltr_20140212_direct 
8. Real talk: http://grantland.com/hollywood-prospectus/rethinking-r-kelly-a-fans-second-thoughts/
http://erinshellman.github.io/ 
real_time_real_talk/

real time real talk

  • 1.
    Erin Shellman, PhD @erinshellman ! Marketing Analytics March 18, 2014 real talk. real-time
  • 2.
  • 5.
    real talk. real-time 1. real-time - what? 2. real talk - look, baby. 3. Nordstrom - real good.
  • 6.
    Big data waslast year, now we want fast data.
  • 8.
    “…Real-Time Analytics is enabling retailers and brands to analyze the right data in real-time to generate solutions that can be implemented across an organization1.”
  • 9.
    What is areal-time system? A real-time system is one that must satisfy bounded response-time constraints or risk severe consequences, including failure2.
  • 11.
    What is real-time analytics?
  • 12.
    In general, real-timeanalytics can be defined as the use of, or the capacity to use, available enterprise data and resources when needed3.
  • 13.
    “It’s about movingwith greater speed toward previously unknown questions, defining new insights, and reducing the time between when an event happens somewhere in the world and someone responds or reacts to that event4.” — Justin Erickson, Cloudera
  • 14.
    “Typically when we’retalking about real-time or near real-time systems, what we mean is architectures that allow you to respond to data as you receive it without necessarily persisting it to a database first.” — John Akred, Accenture
  • 15.
  • 16.
    Williams Sonoma integratesits customer databases with external demographics to created targeted emails. e.g.
  • 17.
    e.g. Macy’s personalizes their catalogs to the point where they could deliver 500,000 unique versions of a single direct mail catalog.
  • 18.
    e.g. Koupon Mediabuilt a real-time dashboard enabling them to identify key attributes of users who participated in their marketing campaigns.
  • 19.
    Are any ofthese use-cases meaningfully real-time?
  • 20.
  • 21.
  • 22.
  • 23.
    1. If youcan’t act on it in real-time, it’s not really real-time.
  • 24.
    1. If youcan’t act on it in real-time, it’s not really real-time.
  • 25.
    1. If youcan’t act on it in real-time, it’s not really real-time. Real Talk.
  • 27.
    “If you havepeople in the loop, it’s not real-time.” — Joe Hellerstein, UC Berkeley
  • 28.
    2. Real-time isnot the same as all-in-one-place.
  • 29.
    2. Real-time isnot the same as all-in-one-place.
  • 30.
    2. Real-time isnot the same as all-in-one-place. Real Talk.
  • 31.
    3. Real-time systemsfail when messages are late.
  • 32.
    3. Real-time systemsfail when messages are late.
  • 33.
    3. Real-time systemsfail when messages are late. Real Talk.
  • 34.
    IBM helped Germanretailer METRO boost customer satisfaction by 18% with intelligent dressing rooms that detect the shopper's current apparel selection and make appropriate recommendations for accessories. e.g.
  • 35.
    E-mail offers fromMicrosoft to engage with Bing are customized at the time of opening. In 200 milliseconds, an offer is assembled based on the most recent customer responses and the available real-time information about the current customer – leading to up to a 70% e.g. increase in conversion rates.
  • 36.
    What’s the difference? • Real-time systems are largely automated, end-to-end. • Building real-time systems requires deep insight, and not so much the other way around. • Real-time analytics is not the push-button solution it’s pitched as.
  • 37.
  • 39.
    Recommendo 2.0 personalizedrecommendations in real-time!
  • 40.
    Batch Real-time Purchasehistory Demographics Category affinity Brand affinity Collaborative Filtering Items Users 4 … … 0 Clicks Logistic Regression Recs Probability Item 1 0.15 Item 2 0.01
  • 41.
    Target Product Product-to-productPersonalized Batch Session-driven Reorder Reorder Best sellers Prefers blazers Clicks red & orange 3 recs per page 1 2 3 4
  • 42.
    What’s an enterpriseto do? • Be disciplined. • Insight before automation. • Fully invest in data technology!
  • 43.
    Being data-driven isa hot phrase: but what does it mean in practice? It’s not about finding analytical fairy dust that you can sprinkle to make everything better, but a deep-rooted approach to operations, innovation and competition. Creating an organization designed for data-informed decision—the experimental enterprise—is the key to competing in a world that’s being rapidly digitized. — Edd Dumbill, Forbes
  • 45.
    Thanks pals! ! Sara Hogenson - Designer, Nordstrom Innovation Lab Scott Jones, Prashanth Nair, Abid Saifee The Advanced Analytics, Personalization, and Mobile Teams Michael Draugelis - Lockheed R. Kelly - Real talk
  • 46.
    Learn more 1.PSFK Future of Retail trend report. 2. Laplante, Phillip A. “Real-time systems design and analysis.” 3. SAP: http://blogs.sap.com/innovation/analytics/real-time-data-analytics-09636 4. Barlow, Mike. “Real-time big data analytics: emerging architecture.” 5. Infochimps: http://www.infochimps.com/resources/koupon-media-case-study/ 6. “Real-Time Retail: Rapid Response to Consumer Demand” from the International Institute for Analytics’ Retail Analytics Research Council. 7. Dumbill, Edd. Forbes. http://www.forbes.com/sites/edddumbill/2014/02/05/the-experimental-enterprise/? imm_mid=0b76e4&cmp=em-strata-na-na-newsltr_20140212_direct 8. Real talk: http://grantland.com/hollywood-prospectus/rethinking-r-kelly-a-fans-second-thoughts/
  • 47.