際際滷

際際滷Share a Scribd company logo

Emg2015

Jeffrey Tsai
Jeffrey Tsai

The document discusses using social media data and customer relationship management (CRM) data to segment customers. It combines Recency, Frequency, Monetary (RFM) scores from both data sources to cluster customers into four groups: 1) high disseminating value, normal shopping value, 2) both shopping & disseminating is low, 3) high shopping value, normal disseminating value, and 4) high shopping value, low disseminating value. Customer targeting and positioning is then performed based on the clusters to identify which social media pages and product categories each group is most interested in. The hybrid analysis approach provides insights into differences between customer groups.

1 of 28
Download to read offline
THE POWER OF CUSTOMER VALUES WITH SOCIAL MEDIA IN THE MARKET SEGMENTATION JeffreyTsai
Since Social media becomes a buzz word, there are many users to . 2
Increased exposure for companies business is the marketers use social media for sale the things. Source: http://www.pagetraf鍖cbuzz.com/learn-companies-connecting-social-media/17235/
Learning from the markets, recognizing the group of customer patterns still a critical job for marketers Because marketers must have the vision to see the value of customers
According to Woodruff (1997) a customer perceived preference for and evaluation of those products attributes, attribute performances, and consequences arising from use that facilitate (or block) achieving the customers goals and purposes in use situations by Woodruff
Segmentation Identifying meaningfully different groups of customers Targeting Selecting which segment to serve Positioning Implementing chosen image and appeal to chosen segment Marketers Customers STP is a familiar strategic approach in Modern Marketing. Segmentation, Targeting and Positioning (STP) Process
Recency Frequency Money When did the customer last make a purchase? How often is the customer making purchases? What is the customers return rate? RFM models is a scoring model and do not explicitly provide a dollar number for customer value. However, RFM are important past purchase variables that should be good predictors of future purchase behavior of customers. (Gupta et al., 2005) RFM Model
We used practical data combining RFM model and the STP strategy with data- mining in CRM & Social Media. CRM Records FB Interactive Records RFM Scores Calculations Social interactive Fanspage CRM shopping category Customer Segmentation Association Rules & Jaccard Coef鍖cient Data Presentation Customer Targeting Modularity Algorithm & Betweenness Centrality Data Preprocessing Customer Positioning Research Process K-means
+ Gathering the 2,456 customers trading records between Jan. and Apr. from CRM database Crawling over 4,000 fanspages interactive records from Facebook Data Gathering
CRM Records FB Interactive records Data Format Facebook_ID Post_time page_id page_name post_id 100000998715544 2015/1/1 152905932107 152905932107_10152635799292108 1786662017 2015/1/1 152905932107 152905932107_10152635799292108 100001469254677 2015/1/1 127628276929 127628276929_10152470414971930 100000802633627 2015/1/1 127628276929 127628276929_10152478522571930 100007917434130 2015/1/1 127628276929 127628276929_10152478522571930 Facebook_ID Money Among Date Product name Product category 100000144441048 840 1 2015/1/1 12 100000199943019 699 1 2015/1/1 6 100000144441048 150 1 2015/1/1 ( .159)2015 100000998715544 238 1 2015/1/1 101 100000998715544 227 1 2015/1/1 ( ) : Connections
https://www.facebook.com/127628276929/posts/10152470414971930
Data Preprocessing Unit: Month CRM Facebook 1st day of next month - trading date E.g. : 2/1 - 1/1 = 31 Accumulating a months BUY frequency E.g. : 1/1 to 1/31 bought 5 books Accumulating a months spending E.g. : Accumulating 5 times spending = 1000 1st day of next month - visiting date Accumulating a months frequency of put the like Accumulating a months interactive CW Pages posts / Accumulating a months total visiting fanspage E.g. : 2/1 - 1/20 = 11 E.g. : 1/1 to 1/31 push the like button 4 times E.g. : 10 CW pages posts / Accumulating100 posts = 0.1
Facebook ID CRM Facebook R F M R F M 1438706957 9.333 9 2783 18 1 0.018 1667846450 27.024 1 587.5 12.333 5.667 0.103 100000294682354 6 5 1553 31 0 0 636098539 3.548 1 1980 31 0 0 100000193167290 18.617 4 774 22 1 0.007 1018548854 5 2 585 4.75 20.25 0.292 The result of Data Preprocessing near end of month Spend money number of books near middle of month visiting weight number of interaction
RFM Scores We based on the 80/20 rules to segment the data into 鍖ve different levels Locating the buy times 1st level score = 3.0 2nd level score = 8.0 3rd level score=13.0 4th level score = 20.0 The range of scores Reader <= 3.0 = 1 3.0 < Reader <= 8.0 = 2 8.0 < Reader <=13.0 = 3 13.0 < Reader <= 20.0 = 4 Reader > 20.0 = 5 The distribution of the RFM scores Scores level CRM Facebook R F M R F M 1 177 2080 215 1431 2245 1629 2 699 306 737 166 146 347 3 768 64 810 319 37 211 4 554 5 535 438 22 179 5 258 1 159 102 6 90 a lot among readers don't read CWs Facebook
The result of RFM Scores Facebook ID CRM Facebook CRM_scores Facebook_scores R F M R F M R F M R F M 1438706957 9.333 9 2783 18 1 0.018 4 3 4 3 1 1 1667846450 27.024 1 587.5 12.333 5.667 0.103 2 1 2 3 1 2 100000294682354 6 5 1553 31 0 0 4 2 4 1 1 1 636098539 3.548 1 1980 31 0 0 5 1 4 1 1 1 100000193167290 18.617 4 774 22 1 0.007 3 2 3 2 1 1 1018548854 5 2 585 4.75 20.25 0.292 4 1 2 4 2 4 They arent active user in each CW Fanspage
Customer segmentation: K-means Num. of observations Clustering 1 428 2 896 3 497 4 635 Clustering result Clustering 1 2 3 4 CRM R 2.9 2.71 3.27 3.29 F 1.18 1.06 1.22 1.34 M 2.92 2.16 2.83 3.89 Facebook R 3.48 1.06 3.74 1.08 F 1.43 1 1.25 1 M 3.84 1.03 1.78 1.05 1. High Disseminating value, normal shopping value 2. Both shopping & disseminating is low 3. High shopping value, normal disseminating value 4. High shopping value, low disseminating value Loyal readers, sometimes buy some books Not loyal customers loyal buyer, sometimes reader from Facebook Loyal Buyers
Shopping RFM Disseminating RFM 1 2 3 4 ?
18 Facebook_ID CRM Facebook Clusters R F M R F M 1438706957 4 3 4 3 1 1 3 1667846450 2 1 2 3 1 2 3 100000294682354 4 2 4 1 1 1 4 636098539 5 1 4 1 1 1 4 100000193167290 3 2 3 2 1 1 4 1018548854 4 1 2 4 2 4 1 Customer Targeting Facebook_ID Clusters Visiting Fanspages (Likes) Buy Categories 1438706957 3 FB1, FB2, FB3 C1, C2 1667846450 3 FB2, FB3, FB5, FB6 C2, C3, C4 100000294682354 4 FB2, FB5, FB6, FB8 C2, C3, C4 636098539 1 FB1, FB4, FB5, FB7 C1, C3, C4, C5 Hybrid the social interactive Fanspage & CRM shopping categories
The customers are concerning the Fanspages Posts 5652 3754 3053 2717 cheers 2212 2149 1362 1257 952 928 877 managertoday 855 766 676 630 616 nexttv 567 552 smart 537 High Disseminating value, normal shopping value Customer Positioning The result of 1st clustering : 428 Design & Stylish Parenting Traveling Complex
Parenting Health & education Cheer Health Magazine Traveling Parenting Illustrator Celebrity Cosmetic Design Business Customer Positioning The result of 1st clustering : 428
The customers are concerning the Fanspages Posts 519 410 tripass 367 322 voguetaiwan 306 275 _crm 247 245 icook 228 213 goodlife 208 184 183 ( ) 182 _crm 164 _crm 162 goodtv 158 157 _crm 145 141 Foundation & Volunteer Political Traveling Magazines New Tech & Entrepreneur Customer Positioning The result of 2nd clustering : 896 Despite both shopping & disseminating is low, they sometimes buy the books from the Internet
Customer Positioning The result of 2nd clustering : 896 Magazine & Health Family & Parenting Education Parenting Cooking Cosmetic
The customers are concerning the Fanspages Posts icook 4986 4380 3411 2861 mamaclub 2584 2490 2362 2347 voguetaiwan 2172 2170 sisy'sworldnews 2093 1918 1801 qqmei 1681 1635 ( ) 1628 cheers 1606 - lessonsfrommovies 1505 tripass 1410 1380 Magazine Parenting & cooking Customer Positioning The result of 3rd clustering : 497 Political Celebrity Womens talk Illustrator High shopping value, normal disseminating value
Customer Positioning The result of 3rd clustering : 497 Parenting & Living Womens talk Financial Health Parenting Cooking Traveling Celebrity Investment Soul Illustrator Health Celebrity
Political New Tech & Entrepreneur Illustrator The customers are concerning the Fanspages Posts _crm 650 gracetw 332 _crm 282 212 _crm 184 vivianhsu 182 _crm 182 icook 175 165 159 158 duncandesign 156 133 133 130 janethsieh 114 yilan 109 pansci 109 2xpeople2 108 Customer Positioning The result of 4th clustering : 635 High shopping value, low disseminating value
Customer Positioning The result of 4th clustering : 635 Investment Stylish Parenting Cooking Cosmetic Celebrity Financial & Magazine Parenting
Conclusions 1. We established a hybrid method to combine the CRM and social media data. 2. We found the difference of each clusters, which the company focuses on.
Thanks for your attention This is our web site: http://leaderboard.ideas.iii.org.tw/home

More Related Content

Emg2015

  • 1. THE POWER OF CUSTOMER VALUES WITH SOCIAL MEDIA IN THE MARKET SEGMENTATION JeffreyTsai
  • 2. Since Social media becomes a buzz word, there are many users to . 2
  • 3. Increased exposure for companies business is the marketers use social media for sale the things. Source: http://www.pagetraf鍖cbuzz.com/learn-companies-connecting-social-media/17235/
  • 4. Learning from the markets, recognizing the group of customer patterns still a critical job for marketers Because marketers must have the vision to see the value of customers
  • 5. According to Woodruff (1997) a customer perceived preference for and evaluation of those products attributes, attribute performances, and consequences arising from use that facilitate (or block) achieving the customers goals and purposes in use situations by Woodruff
  • 6. Segmentation Identifying meaningfully different groups of customers Targeting Selecting which segment to serve Positioning Implementing chosen image and appeal to chosen segment Marketers Customers STP is a familiar strategic approach in Modern Marketing. Segmentation, Targeting and Positioning (STP) Process
  • 7. Recency Frequency Money When did the customer last make a purchase? How often is the customer making purchases? What is the customers return rate? RFM models is a scoring model and do not explicitly provide a dollar number for customer value. However, RFM are important past purchase variables that should be good predictors of future purchase behavior of customers. (Gupta et al., 2005) RFM Model
  • 8. We used practical data combining RFM model and the STP strategy with data- mining in CRM & Social Media. CRM Records FB Interactive Records RFM Scores Calculations Social interactive Fanspage CRM shopping category Customer Segmentation Association Rules & Jaccard Coef鍖cient Data Presentation Customer Targeting Modularity Algorithm & Betweenness Centrality Data Preprocessing Customer Positioning Research Process K-means
  • 9. + Gathering the 2,456 customers trading records between Jan. and Apr. from CRM database Crawling over 4,000 fanspages interactive records from Facebook Data Gathering
  • 10. CRM Records FB Interactive records Data Format Facebook_ID Post_time page_id page_name post_id 100000998715544 2015/1/1 152905932107 152905932107_10152635799292108 1786662017 2015/1/1 152905932107 152905932107_10152635799292108 100001469254677 2015/1/1 127628276929 127628276929_10152470414971930 100000802633627 2015/1/1 127628276929 127628276929_10152478522571930 100007917434130 2015/1/1 127628276929 127628276929_10152478522571930 Facebook_ID Money Among Date Product name Product category 100000144441048 840 1 2015/1/1 12 100000199943019 699 1 2015/1/1 6 100000144441048 150 1 2015/1/1 ( .159)2015 100000998715544 238 1 2015/1/1 101 100000998715544 227 1 2015/1/1 ( ) : Connections
  • 12. Data Preprocessing Unit: Month CRM Facebook 1st day of next month - trading date E.g. : 2/1 - 1/1 = 31 Accumulating a months BUY frequency E.g. : 1/1 to 1/31 bought 5 books Accumulating a months spending E.g. : Accumulating 5 times spending = 1000 1st day of next month - visiting date Accumulating a months frequency of put the like Accumulating a months interactive CW Pages posts / Accumulating a months total visiting fanspage E.g. : 2/1 - 1/20 = 11 E.g. : 1/1 to 1/31 push the like button 4 times E.g. : 10 CW pages posts / Accumulating100 posts = 0.1
  • 13. Facebook ID CRM Facebook R F M R F M 1438706957 9.333 9 2783 18 1 0.018 1667846450 27.024 1 587.5 12.333 5.667 0.103 100000294682354 6 5 1553 31 0 0 636098539 3.548 1 1980 31 0 0 100000193167290 18.617 4 774 22 1 0.007 1018548854 5 2 585 4.75 20.25 0.292 The result of Data Preprocessing near end of month Spend money number of books near middle of month visiting weight number of interaction
  • 14. RFM Scores We based on the 80/20 rules to segment the data into 鍖ve different levels Locating the buy times 1st level score = 3.0 2nd level score = 8.0 3rd level score=13.0 4th level score = 20.0 The range of scores Reader <= 3.0 = 1 3.0 < Reader <= 8.0 = 2 8.0 < Reader <=13.0 = 3 13.0 < Reader <= 20.0 = 4 Reader > 20.0 = 5 The distribution of the RFM scores Scores level CRM Facebook R F M R F M 1 177 2080 215 1431 2245 1629 2 699 306 737 166 146 347 3 768 64 810 319 37 211 4 554 5 535 438 22 179 5 258 1 159 102 6 90 a lot among readers don't read CWs Facebook
  • 15. The result of RFM Scores Facebook ID CRM Facebook CRM_scores Facebook_scores R F M R F M R F M R F M 1438706957 9.333 9 2783 18 1 0.018 4 3 4 3 1 1 1667846450 27.024 1 587.5 12.333 5.667 0.103 2 1 2 3 1 2 100000294682354 6 5 1553 31 0 0 4 2 4 1 1 1 636098539 3.548 1 1980 31 0 0 5 1 4 1 1 1 100000193167290 18.617 4 774 22 1 0.007 3 2 3 2 1 1 1018548854 5 2 585 4.75 20.25 0.292 4 1 2 4 2 4 They arent active user in each CW Fanspage
  • 16. Customer segmentation: K-means Num. of observations Clustering 1 428 2 896 3 497 4 635 Clustering result Clustering 1 2 3 4 CRM R 2.9 2.71 3.27 3.29 F 1.18 1.06 1.22 1.34 M 2.92 2.16 2.83 3.89 Facebook R 3.48 1.06 3.74 1.08 F 1.43 1 1.25 1 M 3.84 1.03 1.78 1.05 1. High Disseminating value, normal shopping value 2. Both shopping & disseminating is low 3. High shopping value, normal disseminating value 4. High shopping value, low disseminating value Loyal readers, sometimes buy some books Not loyal customers loyal buyer, sometimes reader from Facebook Loyal Buyers
  • 18. 18 Facebook_ID CRM Facebook Clusters R F M R F M 1438706957 4 3 4 3 1 1 3 1667846450 2 1 2 3 1 2 3 100000294682354 4 2 4 1 1 1 4 636098539 5 1 4 1 1 1 4 100000193167290 3 2 3 2 1 1 4 1018548854 4 1 2 4 2 4 1 Customer Targeting Facebook_ID Clusters Visiting Fanspages (Likes) Buy Categories 1438706957 3 FB1, FB2, FB3 C1, C2 1667846450 3 FB2, FB3, FB5, FB6 C2, C3, C4 100000294682354 4 FB2, FB5, FB6, FB8 C2, C3, C4 636098539 1 FB1, FB4, FB5, FB7 C1, C3, C4, C5 Hybrid the social interactive Fanspage & CRM shopping categories
  • 19. The customers are concerning the Fanspages Posts 5652 3754 3053 2717 cheers 2212 2149 1362 1257 952 928 877 managertoday 855 766 676 630 616 nexttv 567 552 smart 537 High Disseminating value, normal shopping value Customer Positioning The result of 1st clustering : 428 Design & Stylish Parenting Traveling Complex
  • 21. The customers are concerning the Fanspages Posts 519 410 tripass 367 322 voguetaiwan 306 275 _crm 247 245 icook 228 213 goodlife 208 184 183 ( ) 182 _crm 164 _crm 162 goodtv 158 157 _crm 145 141 Foundation & Volunteer Political Traveling Magazines New Tech & Entrepreneur Customer Positioning The result of 2nd clustering : 896 Despite both shopping & disseminating is low, they sometimes buy the books from the Internet
  • 22. Customer Positioning The result of 2nd clustering : 896 Magazine & Health Family & Parenting Education Parenting Cooking Cosmetic
  • 23. The customers are concerning the Fanspages Posts icook 4986 4380 3411 2861 mamaclub 2584 2490 2362 2347 voguetaiwan 2172 2170 sisy'sworldnews 2093 1918 1801 qqmei 1681 1635 ( ) 1628 cheers 1606 - lessonsfrommovies 1505 tripass 1410 1380 Magazine Parenting & cooking Customer Positioning The result of 3rd clustering : 497 Political Celebrity Womens talk Illustrator High shopping value, normal disseminating value
  • 24. Customer Positioning The result of 3rd clustering : 497 Parenting & Living Womens talk Financial Health Parenting Cooking Traveling Celebrity Investment Soul Illustrator Health Celebrity
  • 25. Political New Tech & Entrepreneur Illustrator The customers are concerning the Fanspages Posts _crm 650 gracetw 332 _crm 282 212 _crm 184 vivianhsu 182 _crm 182 icook 175 165 159 158 duncandesign 156 133 133 130 janethsieh 114 yilan 109 pansci 109 2xpeople2 108 Customer Positioning The result of 4th clustering : 635 High shopping value, low disseminating value
  • 26. Customer Positioning The result of 4th clustering : 635 Investment Stylish Parenting Cooking Cosmetic Celebrity Financial & Magazine Parenting
  • 27. Conclusions 1. We established a hybrid method to combine the CRM and social media data. 2. We found the difference of each clusters, which the company focuses on.
  • 28. Thanks for your attention This is our web site: http://leaderboard.ideas.iii.org.tw/home
AboutCopyright
息 2025 際際滷