Netsci2010 social networks for marketers


Published on

Published in: Business
  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide
  • 05/15/10
  • 05/15/10
  • presentation title
  • presentation title
  • presentation title
  • 05/15/10
  • 05/15/10
  • Analytical CRM - ppt for Consumer Marketing Author - Simon Dudley
  • Netsci2010 social networks for marketers

    1. 1. The marketers point of view Applications and challenges for marketers using call network analysis 11 May 2010, Matej Karpisek Presented at: NetMob 2010 (satelite of NetSci 2010)
    2. 2. The 15’ path <ul><li>Who is ENQIO </li></ul><ul><li>Application 1: Viral marketing </li></ul><ul><li>Application 2 : Community marketing </li></ul><ul><li>Application 3 : data completion </li></ul><ul><li>Application 4 : Rotational churn </li></ul>
    3. 3. WHO IS ENQIO? ENQIO | <footer> 15/05/10 | Page
    4. 4. Company <ul><li>Offering our Customers “High Value Solutions” to boost Revenue & Margin </li></ul><ul><li>A multi-disciplinary team with a unique mix of skills </li></ul><ul><ul><li>Data Mgt experts, Marketers, statisticians, business intelligence and data mining experts, mathematicians, computer scientists, economists, sociologists, trouble-shooters and value seekers. </li></ul></ul><ul><ul><li>Cross-competence skills allow fast switching between top down and bottom up </li></ul></ul><ul><li>Technical Infrastructure designed for “large data volume processing” </li></ul><ul><ul><li>Solid information platform; highly secured </li></ul></ul><ul><ul><li>+ 50TB storage and processing of CDR’s </li></ul></ul><ul><li>ISO 9001:2008 Certified </li></ul><ul><ul><li>Focus on “high quality delivery of products” </li></ul></ul><ul><ul><li>Clearance to work for international government /bodies </li></ul></ul><ul><li>Experienced Management team </li></ul><ul><ul><li>Telecom, Media and Banking </li></ul></ul><ul><ul><li>Experience in “Solutions Building” </li></ul></ul><ul><ul><li>International experience </li></ul></ul><ul><li>HQ in Belgium </li></ul><ul><ul><li>Working for more than 40 countries </li></ul></ul>15/05/10 | Page ENQIO | mobile marketing forum
    5. 5. Vision <ul><li>Information and knowledge isn’t enough </li></ul><ul><li>Without insight Data is worthless </li></ul><ul><li>Success = Understand </li></ul><ul><ul><li>its target market </li></ul></ul><ul><ul><li>its range of needs and potential </li></ul></ul><ul><ul><li>how to best act upon this information </li></ul></ul>15/05/10 | Page ENQIO | mobile marketing forum
    6. 6. Strategy 15/05/10 | Page “ Making Data Work” by providing solutions that generate a measurable higher return on information. (ROi²) ENQIO | mobile marketing forum
    7. 7. Social networks @ ENQIO BE, June 2006: Exceptional successful campaign NL, Apr 2007, MMS& VAS upsell campaigns NL, May 2006: Start of SNA in Netherlands NL, July 2006: Net promoter score analysis PT, Oct 2006: Start of SNA in Portugal ES, Aug 2006: Benchmarking of ENQIO social leaders PT, Oct 2007: Launch of get/keep/increase campaigns SK, May 2007, Viral marketing in Slovakia FR, Jan 2008, Viral marketing research in France BE , Nov 2004 : Analysis of contact circle value BE, Nov 2005: R&D , social leader, communities BE, Aug 2009, Next gen SNA initiative CA, Jan 2008, Start of VM campaigns 100 + VM campaigns executed BE, Aug 2008, iPhone analysis
    8. 8. Social network analytics @ ENQIO ENQIO Social Network users Successful SNA applications <ul><li>Viral stimulation of MMS usage </li></ul><ul><li>Introduction of new tariff plans through social leaders </li></ul><ul><li>Successful introduction of a high value handset </li></ul><ul><li>Up sell for prepaid VAS services with SNA </li></ul><ul><li>.... </li></ul><ul><li>“ Our customers executed more than 200 successful campaigns based on ENQIO SNA analytics” </li></ul>SNA offering <ul><li>Solutions for </li></ul><ul><li>Social leader identification </li></ul><ul><li>Detection of rotational churn </li></ul><ul><li>general SNA analytics </li></ul><ul><li>Applied in </li></ul><ul><li>Get, keep, increase actions </li></ul><ul><li>Propensity models </li></ul><ul><li>Customer value calculations </li></ul><ul><li>Review of acquisition and retention budgets </li></ul><ul><li>Experienced team </li></ul><ul><li>5+ years of experience with SNA in telco </li></ul><ul><li>Proven success in past projects </li></ul>
    9. 9. VIRAL MARKETING? ENQIO | <footer> 15/05/10 | Page
    10. 10. Day 0: blogger publishes a movie : How to pick a Kryptonite lock with a pen <ul><li>The sales of Kryptonite locks collapses </li></ul><ul><li>Class action suit </li></ul><ul><li>Recall action for the locks </li></ul>Day 1: A news site on biking picks up the Blog post and refers to it “ Several readers try it at home and confirm the story” a year later: Kryptonite starts to regain shelf space in the stores How many are unlikely to ever again purchase a Kryptonite product as a result of this? Question of a customer: day 5: The NY-times publishes the story Kryptonite also picks up the story and looks for a solution... It would take 3 weeks to organize a call-back actions + change of manufacturing process
    11. 11. The strength of WOM Source: MEC MediaLab
    12. 12. Word of mouth, the extra direct marketing channel <ul><li>“ Viral marketing uses the power of social interactions to spread the message and influence the behavior of customers” </li></ul><ul><li>Above the line </li></ul><ul><li>Direct mail </li></ul><ul><li>Outbound call </li></ul><ul><li>E-mail </li></ul><ul><li>In shop communication </li></ul><ul><li>.... </li></ul>“ The message comes from a Telco” “ The message is spread by other customer” <ul><li>Viral messages </li></ul><ul><li>Usage stimulation with call me back </li></ul><ul><li>Influence by contacts </li></ul><ul><li>... </li></ul>Influence / impact LOW HIGH  Customers trust opinions of other customers more than direct communications  Less customers need to be contacted to reach a large group of people/customers  Customer evangelists allow a faster spread of new technologies/services “ An active customer recommendation leads to a conversion rate up to 85%” Boosted response on DM campaign from 5% to 35%
    13. 13. multichannel communication CRM Toolkit Bill Insert, Bill message, Bill envelope. Bill Outbound Email campaign, SMS, MMS Video, Direct Mail. Segmented Email Segmented Email and associated segmented web landing pages. Offers Offer developments, IEW treatments offers. Contact Centre Contact centre script (in out bound calling) Service support, FAB Briefing packs and agent ref material, promotional items. Two way SMS Two way SMS to request international roaming Video Video promo, viral series video utube, Video service support BTL Activity Sales promotion techniques for target segment, Sponsorship opportunities, PR concepts for promoting approach, Competition prizes, Member get member WOM Steer the WOM by targeting social leaders : Increase Word Of mouth , steer the content (right message)
    14. 14. MMS campaign Prepaid social leaders not using MMS yet with MMS enabled handsets, contacted via SMS Target 100 free MMS per month during 3 consecutive months In addition, possibility to fwd offer to friends Offer Communication 9.5 times above expectations 2 times above expectations 5 times above expectations
    15. 15. Happy Hour campaign Pay as you go social leaders with on-net voice spending lower than €4 per month Target 1 month free Happy Hour (VAS on top of tariffplan) fee of 5 euro will be applied Offer Communication Direct SMS: 1.2% e-mail: 2.4% Results  SMS: 2.000 leaders  Email: 286 leaders Response rate Improvement over previous campaigns On top of direct, 3-4 times higher conversion in circle Viral : 4.3%
    16. 16. Looks nice …. <ul><li>But it does not always work </li></ul>Voicemail activation  no spread Acquisition & retention campaigns  no success
    17. 17. Virality of products ENQIO | <footer> 15/05/10 | Page Leaders boost spread Leaders block spread Low viral spread High viral spread ADSL Blackberry Mobile data Content MMS High voice Low voice Low SMS Banxafe Convergence & co-branding Mobile services Customer behaviour
    18. 18. To learn... <ul><li>How can we make viral marketing and churn/acquisition work </li></ul>ENQIO | <footer> 15/05/10 | Page
    19. 19. COMMUNITY MARKETING ENQIO | <footer> 15/05/10 | Page
    20. 20. Impact on dongle adoption when convincing a social leader No social leader in the circle Social leader without dongle in the circle Social leader with dongle in the circle One of my friends has a dongle Two of my friends have a dongle Three of my friends a dongle No one in my circle has a dongle Effect of in the circle of converting a social leader My average dongle adoption rate ENQIO | netmob 2010
    21. 21. Importance of communities <ul><li>School classes, work colleagues, family, groups of friends… </li></ul><ul><ul><li>The Orange customer base can be treated as one group or as a collection of communities. </li></ul></ul><ul><li>Breakdown of the product diffusion curve per community for new products </li></ul>Spread per community Spread in customer base innovators Early adopters Early majority Late majority laggards Customer Advertising, Sales promo Word to mouth “The Buzz” Direct marketing on “Seeds” Equivalent views on customer base Viral or Buzz marketing approach ENQIO | netmob 2010
    22. 22. The customer base, not one large community Differences in language Geographical spread Each dot represents a customer. The distance between 2 dots represents the distance/closeness in the social network. Afterward the dots are coloured based on the language or geographical location ENQIO | netmob 2010
    23. 23. When do people call Low users: 9AM till 4PM , 7days a week Business Business hours Youngsters evenings, 7days a week and weekend nights Time of day/week when customers from different segments contact each other ENQIO | netmob 2010
    24. 24. Events in communities ENQIO | netmob 2010 15/05/10 | Page First MMS usage Churn Handset switch
    25. 25. Which algorithm to calculate these communities? <ul><li>Simple allocation rules (iterative process) </li></ul><ul><ul><li>Person A is better connected to community X...  move </li></ul></ul><ul><li>Breakdown of the network </li></ul><ul><ul><li>Betweenness metrics </li></ul></ul><ul><li>All algorithms have the same issues </li></ul><ul><li>High variance of results </li></ul><ul><ul><li>With different parameters </li></ul></ul><ul><ul><li>With different algorithms </li></ul></ul><ul><li>Difficult to interprete the results </li></ul><ul><ul><li>What does it mean to belong to community X </li></ul></ul><ul><ul><li>What is the practical difference between different approaches </li></ul></ul>15/05/10 | Page ENQIO | netmob 2010
    27. 27. What is rotational churn? 15/05/10 | Page ENQIO | netmob 2010
    28. 28. Spinning / rotational churn <ul><li>Spinners are customers that stop using their SIM card and get a new one (driven by marketing actions, incorrect info at POS, Lost SIM,...) </li></ul><ul><li>Main Issues : </li></ul><ul><li>ARPU drop of spinners during ‘spin’ </li></ul><ul><li>Unwanted SAC and SRC costs </li></ul><ul><li>Incorrect understanding of ‘real’ churn </li></ul>15/05/10 | Page Distribution of new Prepaid activations ENQIO | netmob 2010
    29. 29. Spinning / rotational churn <ul><li>Differentiated approach for </li></ul><ul><ul><li>Spinners </li></ul></ul><ul><ul><li>Understand gaps in communication (e.g. change of MSISDN does not require a new card) </li></ul></ul><ul><ul><li>Upgrade paths where customers are spinning today to upgrade </li></ul></ul><ul><ul><li>Remove from propensity models </li></ul></ul><ul><ul><li>Channels promoting spinners </li></ul></ul><ul><ul><li>Fake acquisitions </li></ul></ul><ul><ul><li>Tag actions that attract spinners </li></ul></ul><ul><ul><li>Remove from propensity models </li></ul></ul><ul><ul><li>Real churn/acquisition </li></ul></ul><ul><ul><li>This is the target of propensity models </li></ul></ul><ul><li>Based on spinners model </li></ul><ul><li>Input towards </li></ul><ul><ul><li>LTV calculations </li></ul></ul><ul><ul><li>Reconnect spinner pairs to look at total value and total lifetime </li></ul></ul><ul><ul><li>Are FA profitable? </li></ul></ul><ul><ul><li>Correction of acquisition and retention costs </li></ul></ul><ul><li>Remove welcome call for spinners </li></ul>Value of spinner identification
    30. 30. Variable and event decomposition Common contacts on the FIX net (low expected variation) Common contacts mobile off net (Rotational churn exists here but with different promo’s) Common contacts mobile on net (Rotational churn makes the contacts also less stable) Are they the same person? Do they have the same contacts? Do they call from the same places? Sim A Sim B Cell A Cell X Cell F Most used cell 2nd cell 3th cell Cell A Cell Z Cell X  Similarity score ENQIO | netmob 2010
    31. 31. Equal contacts (v’s calculation) Candidate Contract A Candidate Contract B Fix line On-net post On-net pre Other mobile Input data : CDR data for the customer base For voice/SMS/(MMS) “ Ideal case 3 months before and 3 months after” Short numbers  Importance of call direction needs to be verified ENQIO | netmob 2010
    32. 32. Rotational churn remains a challenge <ul><li>No target to model on </li></ul><ul><li>Current models are too slow for operational use </li></ul><ul><li>(they are useful for strategic practices) </li></ul>15/05/10 | Page ENQIO | netmob 2010
    34. 34. Stability 15/05/10 | Page ENQIO | netmob 2010
    35. 35. Stability questions 15/05/10 | Page ENQIO | netmob 2010 Distribution over different connection times over 1 year of time Only a small fraction of contact is strong and stable (core)
    36. 36. Understand Social networks <ul><li>Calling circle composition </li></ul>15/05/10 | Page 22% of the calling circle consists of Connections with incoming and outgoing traffic 39% of the calling circle consists of Connections with only outgoing traffic 39% of the calling circle consists of Connections with only incoming traffic 22% IN + OUT 39% IN 39% OUT ENQIO | netmob 2010
    37. 37. Grouping of ethnicities Understanding ethnicity adds value to customer understanding However privacy regulation prohibit the use of this information All customers No name known Uncharacteristic name Characteristic name Asian Name Northern European Western European Southern European Eastern European Island Other 28.5% 23.0% 48.5% chinese indian indonesian hindoe japanese kazakh korean thai vietnamese danish finnish icelandic swedish norwegian breton french frisian galician german dutch provencial basque catalan spanish italian greek corsican portuguese latvian lithuanian macedonian romanian russian serbian_and_croatian slovene albanian ukrainian polish estonian czech_and_slovak bulgarian english welsh scottish modern_english native_american cornish traditional_english irish african armenian azerbaijani esperanto hawaiian igbo jewish khmer kurdish manx maori yoruba  
    38. 38. Sometimes controvertional <ul><li>Net promoter score </li></ul><ul><li> promise best metrics to track performance of your business (Reicheld) </li></ul><ul><li> detail analysis with social networks + MR </li></ul><ul><li>no impact/correlations with growth </li></ul>ENQIO | <footer> 15/05/10 | Page
    39. 39. Missing data 15/05/10 | Page OFF NET Postpaid Prepaid ENQIO | netmob 2010 Offnet connections are not visible in the graph... How to deal with this?
    40. 41. Get the word out...