Big Social Data
                      From Big Data to Smart Data
                                                    Prinzie Anita


solutions-2 Belgium    www.solutions2.be             +32(0)9 242 50 40
Solutions-2 London     www.solutions2.co.uk          +44 (0)20 7608 9300
OR


50,000   100,000 125,000 150,000
OR


650,000   100,000 48 hours   28,000
http://datasift.com/stream/13991/mcdo
nalds#app1-preview
ORGANIC DATA          DESIGNED DATA




  CREATED TO GAIN SPECIFIC INSIGHTS
           DATA RICHNESS


                       Based on Bob Groves, former US Census director
BIG DATA   SMART DATA
Define business
   objective
LISTENING                MONITORING




Brand perceptions        Detecting crisis

Brand positioning        Complaints & malfunctions

Consumer segments        New product launch

Media campaign success   ...

...                             Based on Jasper Snyder Converseon
Identify relevant data
BUSINESS
                      OBJECTIVE

LISTENING PURPOSE                   MONITORING PURPOSE




                    RELEVANT DATA

                Who?
                Which platforms?
                Which conversations?
                How long?
Who
Fit with business objective
                         IPhone 5 UK launch success (Vision Critical)
                         “Should we listen to FB conversations
                         of people not wanting the IPhone 5?”


                                      Battery
                                8                3
                                                44
                               65                           Negative
                                                            Neutral
                                                            Positive
                                                53
                               27

                                All        Want IPhone 5


                                                                       11
Who & Which platforms
Profile on channel usage & engagement




                          15-24   25-34   35-54   55-99




                              Actively engaging on
                               FB, Twitter, blogs?
Which conversations
Find the most relevant ones
              B2B Software Adoption Journey

              Focus groups with B2B customers



               Dictionary of typical actions during different phases
               of the software adoption process



               Scoring all Twitter/blog conversations
               on the software adoption phases.


               Software adoption journey-conversations


                                                                       13
How long?
Find relevant time window

                       Natural time window


                       ‘Enough conversations’ time
                       window


                       New product launch: 90 days
                       before and after (Microsoft)


                                                      14
Clean &
Preprocess
Clean and Preprocess
Keep goal in mind


                    Keep emoticons and markup
                    for detecting crisis



                    Context specific normalization
                    and annotation (e.g. Convey
                    API)



                                                  16
Analyse with
objective
in mind
MONITORING PURPOSE        SENTIMENT ANALYSIS



Detecting crisis         Detecting gradations of
                         negative and evolutions
Identifying complaints   Detecting gradations of
and malfunctions         negative


Monitoring response to   Detecting gradations of
new product launch       positive and negative

                                                   18
Remedies

       Correct
        80%
                                      Cost-sensitive learning


                                      Undersampling
                                      of neutral class




                  90%
                                      Use recall, precision
                                      and F1 measure to
  7%                         3%       evaluate model
Positive         Neutral   Negative
                                                                19
Validate
results
Validate social media results



              Denali    The evolution of the Microsoft

                        software adoption index did

                 O365   follow known success/failure

                        trends for past software

                        launches.




                                                         21
BIG DATA


             Define business objective

             Identify relevant data

             Clean & preprocess

             Analyse with objective in mind

             Validate results



SMART DATA
Anita Prinzie
                anita@solutions2.be
                @AnitaPrinzie




        If you have any

risingquestions
References



   http://www.domo.com/blog/2012/06/how-much-data-is-created-
    every-minute/
   Snyder, J. (2012), Enriching Social Data for Market Research
    Converseon, New MR Webinar, Social Media Research, October
    9th 2012.
   Woolmer, J. (2012), Is it real? Using conventional research to
    validate and quantify social media findings, Vision Critical, New
    MR Webinar, Social Media Research, October 9th 2012.




                                                                        24

From Big Social Data to Smart Social Data

  • 1.
    Big Social Data From Big Data to Smart Data Prinzie Anita solutions-2 Belgium www.solutions2.be +32(0)9 242 50 40 Solutions-2 London www.solutions2.co.uk +44 (0)20 7608 9300
  • 2.
    OR 50,000 100,000 125,000 150,000
  • 3.
    OR 650,000 100,000 48 hours 28,000
  • 4.
  • 5.
    ORGANIC DATA DESIGNED DATA CREATED TO GAIN SPECIFIC INSIGHTS DATA RICHNESS Based on Bob Groves, former US Census director
  • 6.
    BIG DATA SMART DATA
  • 7.
  • 8.
    LISTENING MONITORING Brand perceptions Detecting crisis Brand positioning Complaints & malfunctions Consumer segments New product launch Media campaign success ... ... Based on Jasper Snyder Converseon
  • 9.
  • 10.
    BUSINESS OBJECTIVE LISTENING PURPOSE MONITORING PURPOSE RELEVANT DATA Who? Which platforms? Which conversations? How long?
  • 11.
    Who Fit with businessobjective IPhone 5 UK launch success (Vision Critical) “Should we listen to FB conversations of people not wanting the IPhone 5?” Battery 8 3 44 65 Negative Neutral Positive 53 27 All Want IPhone 5 11
  • 12.
    Who & Whichplatforms Profile on channel usage & engagement 15-24 25-34 35-54 55-99 Actively engaging on FB, Twitter, blogs?
  • 13.
    Which conversations Find themost relevant ones B2B Software Adoption Journey Focus groups with B2B customers Dictionary of typical actions during different phases of the software adoption process Scoring all Twitter/blog conversations on the software adoption phases. Software adoption journey-conversations 13
  • 14.
    How long? Find relevanttime window Natural time window ‘Enough conversations’ time window New product launch: 90 days before and after (Microsoft) 14
  • 15.
  • 16.
    Clean and Preprocess Keepgoal in mind Keep emoticons and markup for detecting crisis Context specific normalization and annotation (e.g. Convey API) 16
  • 17.
  • 18.
    MONITORING PURPOSE SENTIMENT ANALYSIS Detecting crisis Detecting gradations of negative and evolutions Identifying complaints Detecting gradations of and malfunctions negative Monitoring response to Detecting gradations of new product launch positive and negative 18
  • 19.
    Remedies Correct 80% Cost-sensitive learning Undersampling of neutral class 90% Use recall, precision and F1 measure to 7% 3% evaluate model Positive Neutral Negative 19
  • 20.
  • 21.
    Validate social mediaresults Denali The evolution of the Microsoft software adoption index did O365 follow known success/failure trends for past software launches. 21
  • 22.
    BIG DATA Define business objective Identify relevant data Clean & preprocess Analyse with objective in mind Validate results SMART DATA
  • 23.
    Anita Prinzie anita@solutions2.be @AnitaPrinzie If you have any risingquestions
  • 24.
    References  http://www.domo.com/blog/2012/06/how-much-data-is-created- every-minute/  Snyder, J. (2012), Enriching Social Data for Market Research Converseon, New MR Webinar, Social Media Research, October 9th 2012.  Woolmer, J. (2012), Is it real? Using conventional research to validate and quantify social media findings, Vision Critical, New MR Webinar, Social Media Research, October 9th 2012. 24

Editor's Notes

  • #23 So in a nutshell. To convert big social data into smart social data you follow a five step process.In the first step, you clearly define your business objective so you know whether you need social media listening or social media monitoring.In the second step, you identify the “relevant” social media data.In the third step, you clean and preprocess this “relevant” social media data perhaps with .In the fourth step, you analyse your cleaned “relevant” social media data with your business objective in mind.In the fifth step, you validate your social media results.So, now you have smart social data!