+ All Categories
Home > Documents > Presentazione Catarina Sismeiro.pdf

Presentazione Catarina Sismeiro.pdf

Date post: 26-Feb-2018
Category:
Upload: anonymous-5l5grcup
View: 218 times
Download: 0 times
Share this document with a friend
32
BI G D AT A = BI G DISRUPTION? NUOVE FRONTIERE DEL BRAND ENTERTAINMENT Catarina Sismeiro Imperial College London Associate Professor of Marketing
Transcript

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 1/36

BIG DATA = BIGDISRUPTION?

NUOVE FRONTIERE DELBRAND ENTERTAINMENT 

Catarina SismeiroImperial College London

Associate Professor ofMarketing

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 2/36

OVERVIEW

1. Big data

2. Branded Entertainment and how itcan benefit from big data analytics

3. Applications and examples

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 3/36

BIG DATA?  ATA THAT REQUIRES

NEW SKILLS AND A

W WAY OF LOOKING AT STORAGE ANDPROCESSING

“Big data describes datasets that are so lar

complex, or rapidly changing that they push the v

limits of our analytical capability. It's a subjective te

What seems ‘big’ today may seem modest in a f

years when our analytic capacity has improve

 Joel Gurin, author of Open Data Now

What are the typical challenges?!  High Volume

High Velocity

!  Extremely Unstructured

Not Just associated with internet(“Old Data” can be very “big”!)

New data available in a world of “all-things-digital”(e.g., text, image, voice/sound, and vide

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 4/36

 PROJECT

 A BOLD IDEA TO

SHOWCASE THEOWER OF BIG DATA ANALYTICS

Black Box… the system does not try to understa

“why” but wants only to predict

When something changes in the market, predicti

can go wary.More than just Big Data… need Intelligent Data

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 5/36

NOW CLOSED

Thank you for stopping by.

Google Flu Trends andDengue Trends are no

publishing 

current estimatesand Dengue fever based on

patterns. […] It is still early d

nowcasting and similar tounderstanding the spread of d

like flu and dengue – we'reto see what comes next. […]

Sincerely,The Google Flu and Dengue

Team.

https://www.google.org/flutabout/ 

PROJECT A BOLD IDEA TO

SHOWCASE THEOWER OF BIG DATA ANALYTICS

Black Box… the system does not try to understand

“why” but wants only to predict

When something changes in the market, predicti

can go wary.More than just Big Data… need Intelligent Data

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 6/36

BRANDENTERTAINMENT

GREATOPPORTUNITIES ANDGREAT CHALLENGES

IN TODAY’SETWORKED SOCIETY

Not new… but new opportunities for content creati

and distribution

Power of amplification  and serendipity of soc

networks (emerged more as a platform of u

distributed content than one of user generated conte

Great potential and but also significant risks

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 7/36

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 8/36

HOW CAN BIG DATA ANALYTICS HELP

BRANDENTERTAINMENT?

Start simple and reuse existing tools and analyses (

of the gains from the initial 20% of effort)

Think of the relevant unit of analysis

(e.g., if I make decisions weekly should I have an h

demand analysis?)

Combine methods

!  Discovery and Sensing

!  Measurement and Prediction

!  Targeting and Personalization

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 9/36

 APPLICATIONPREDICTING VIRALITY ANDHELPING DESIGNUSING TEXT ANALYSIS

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 10/36

TEXT ANALYSIS:SIMPLE EXAMPLE

Calzedonia Ocean Girls SonoStupende!!!” 

After stemming and after removing

extremely common and extremelyuncommon words, retained only:

calzedonia, ocean, girls,

Stupende

Encode as TextVector (word_id : count)

(5:1,7:1,142:1,846:5)

Create word dummies to use in models

Message Word 1 Word 2 … Wor

1 1 0 0

2 0 0 1

… …

N 1 1 0

0! Word not present1! Word present

Stemming example: eater, eating ! eat

This post would include one mention tomonitored brand. The same can be done

longer texts (e.g., article) like in the vira

study

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 11/36

 BEYOND VOLUM

 SENTIMENT A

MEANING 

The apparentincrease in braninterest was no

necessarily positiSentiment Analy

and Content Anal(i.e., what is said

if it is positive o

negative) is ver important

But remember t 

 ABERCROMBIE AND

FITCH

“FITCH”

GOOGLERENDS DATA

“FITCHTHE

HOMELESS”

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 12/36

 FINDINGS

CONTENT VIRALIT

DEPENDS ON:

Position of the content(top of the list)

!  Practical utility of post

!  EMOTIONS

Content associated witharousal positive (awe, surand humour) or negati

(anger or anxiety) emotimore viral

Content that evokes lo

arousal, or deactivatinemotions (sadness) is less

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 13/36

  ANALYSIS

MEASURING BRANDSTRENGTH (IFEGY ATLAS)

Free or commercisoftware perform t

mining for differelanguages

Data can come fr blogs, online

conversations in somedia, news, ademails, commerci

offers, reportsThink of the potent

What are peopldiscussing afterwatching the lasepisode of RDS

 Academy? Discover ideas for

next episode? Discwhat worked and

did not?What content on

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 14/36

 APPLICATIONTARGETING MOBILE USERS WITHMULTIMEDIA MESSAGES: POWEROF TEXT AND IMAGE

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 15/36

15

lemobile phone company wishes to target

stomers with alternative messagesessages include offers from manytegories and combine TEXT + IMAGE+

RICE

ifficulty in learning using traditional testing

ethods due to the large scale of offers ande reduced time to learn (e.g., most expireuickly not enough testing opportunities)

Solution –  Statistical methods (SVR), text mining (sim

word count), and automatic processing ofimages (Textons) to predict performancedifferent users

 –  Given limited testing possibilities, test onlmost promising offers as per the model

 – 

Practical issues of deployment that lead tadaptive system (control vs. testing group

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 16/36

 digital representation of images:ixels (a structure that provides a

location”; could be similar to a Cartesianrid though alternatives exist)

olour encoding (e.g., RGB)

m their digital representation we canprocess images using a variety of

rs, extract features (e.g, extract textextract components, determine howuch vegetation, or how much “skin”)nd, considering their overall look atl, classify and group similar images

COMPOSED INTO

LOUR COMPONENTS

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 17/36

MAGE ALYSIS

:IXELS ANDILTERS

Examples of Filtered Images

There are many different filters

some allow the classification on

overall look and feel of the imaothers try to extract components

the image or identify edges a

ridges

red values are then

ed (e.g., K-means) and

ntroid of each cluster

ecomes a visual

“word” (Texton)

then describe each

e with respect of how

f each “visual word” it

contains

lso possible to extractic image components

Pixel 

Each image is made of pixels

(similar to a coordinate, a location)

associated to a specific value (e.g.,

in RGB colour encoding)

For simplicity we are consideringhere a rectangular grid (this is not

the only possibility)

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 18/36

 AUTOMATIC

PROCESSING OFVISUAL

CONTENT

 Automatic imageclustering works verywell in finding similar

types of images

The type of imagebeing used can then

be entered as avariable in models of

performance

Or develop system topredict potentialoutcome

many spetools anddifferent w

of

implementimage min 

Some ar “supervisand requi

some huinput anothers acomplete

unsupervi 

The 80/20

applies alhere(start sim

 

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 19/36

1.0

1.1

1.2

1.3

1.4

1.5

10 15 20 25 30 35 40 45 50 55

Price

Price and T

Price, Texto

Text

   R  e  v  e  n  u  e

   L   i   f  t

System Testing Constraints

RESULTSrmation on images and

t help to significantlyprove performance

e methodology can belied to devise a systemest different messages

ormance metrics couldde clicks, views, shares

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 20/36

 APPLICATIONREAL TIME MARKETING: SENSING AND MONITORING

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 21/36

 “THE STRATEGY A

PRACTICE OFREACTING WITH

IMMEDIACY IN DIGI

CHANNELS TOEXTERNAL EVENT AND TRIGGERS”

OUTCOMES*

81% increased customengagement

73% improved customexperiences

59% increased conversi

rates

Challenges: identifyinopportunities that are

adequate for your target

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 22/36

 

REACTION

Sensing can be done

humans or by automasystems

Sensing systems can ron conversation dat

among the targetconsumers of your pro

(specialized) and setautomatic warnings

Creative can be testusing automatic syste

similar to the one

developed for the previexample relying on im

and text

Sensing also useful for

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 23/36

 APPLICATIONWHAT IS THE VALUE OF A SOCIALMEDIA CONVERSATION FOR ABRAND?

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 24/36

24

lemsearch has shown that companies canfectively stimulate online conversations 

rough innovative campaigns, and that thesenversations increase product sales

ow much to invest in these activities dependsn much conversations are worth in terms ofvenue they generate

ow much is a conversation in different social

edia worth for soft drinks brands that sellostly offline?

Solution –  Collect data on online conversations and

determine when a brand is mentioned

 – 

Link conversation data to offline sales dataperiod) and account for the effect of adver(paid media)

 –  Use statistical models to determine how deis stimulated by conversations

 –  Simulate the value of online conversations uthe estimated model

 –  18 brands, 19 months, 12 US markets

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 25/36

0

510

15

20

25

30

35

40

Facebook Blog/Discussion

Boards

Twitte

Revenue Change Due to 1,000Additional Conversations Across All

Brands

(in US Dollars)

Facebook Blog/Discussion

Boards

Twitter

Average MonthlyConversations Mentioning a Brand

(in ‘000s)

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 26/36

0

10

20

30

40

50

60

Facebook Blog/Discussion Boards Twitter

FOR ALL BRANDS… B AND SMA

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 27/36

BIG DATA ANALYTICSCAN INDEED HELP

BRANDENTERTAINMENT

Remember…

Start simple and reuse existing tools and analyses (

of the gains from the initial 20% of effort)

Think of the relevant unit of analysis (more than big

it is intelligent data!)

Combine methods

How can we benefit?

!

 

Discovery and Sensing!

 

Measurement and Prediction

Targeting and Personalization

Consider variety of dataBe creative in data structures and

analyses

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 28/36

THANK YOU

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 29/36

New YorkPublic

LibrarySeptember

2011Flash Mob

Partners &

Spade

MarketDisruption

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 30/36

Warby Parker

Searches

WHAT WORKS?THE CASE OF

WARBY PARKER

Lets do some analysis with Google trends… over tim

the company has grown in terms of interest in the

market. We notice some significant spikes…

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 31/36

Google Trends: Searches

for Warby Parker in 2011

tor Traffic to Website

rce: Annual Report 2011

INTEGRATIN AN

INTERACTI

NDIFFERECHANNE

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 32/36

NEW VERSUSTRADITIONAL

Internet is growing fast and new media is gaining

importance

Traditional media is still huge Opportunities exist to

take advantage of their integration and interaction

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 33/36

Decompose the images i

different “layers” (using fil

Create a Dictionary of Vi

Words or Textons = Centroipixel clusters

Even if the system has ne

seen an image before, it

process it using the filtersthen assign each pixel to

closest cluster and deter

how many of the Textons (

words) we have in the im

Final result is a distributioTextons for each pictur

TRAININGIMAGES

 ALLGESVENNEW

ES,

VEREENORE

)

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 34/36

WHY?

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 35/36

 APPLICATION

SOCIAL MEDIA AND ONLINECONTENT CONSUMPTION:PREDICTION AND SOCIALINFLUENCE

7/25/2019 Presentazione Catarina Sismeiro.pdf

http://slidepdf.com/reader/full/presentazione-catarina-sismeiropdf 36/36

  News Sites andFacebook

PREDICTING WITH SOCIAL

MEDIA

36

!"

$!"

%!"

&!"

'!"

(!"

)!"

*!"

+!"

,!"

$!!"

!" $!" %!" &!" '!" (!" )!" *!" +!" ,!" $!!"

'( -./0.. 123.

4511 467.899:

;6<. =9-.1

>9 467.899:

?@3 47.899:

33% improvement in

Page Views prediction

over a model that was

performing alreadyextremely well

Improvement came

mostly from the

information of what

Facebook friends wereviewing at the news site

Predicting

site visit


Recommended