Upgrade to Pro — share decks privately, control downloads, hide ads and more …

LINE TECHPULSE 2022 - AI Solutions of MarTech in LINE

LINE TECHPULSE 2022 - AI Solutions of MarTech in LINE

AI Solutions of MarTech in LINE by Johnson Wu / Charlie Tang / Nina Cheng / TW Data Dev @ LINE TECHPULSE 2022 https://techpulse.line.me/

2102a6b8760bd6f57f672805723dd83a?s=128

LINE Developers Taiwan
PRO

January 21, 2022
Tweet

More Decks by LINE Developers Taiwan

Other Decks in Programming

Transcript

  1. Johnson Wu / Charlie Tang / Nina Cheng / TW

    Data Dev AI Solutions of MarTech in LINE 
  2. Tonight, I’ll be eating Martech(?) ….

  3. Most Common Marketing Challenges

  4. Most Common Marketing Challenges How can I get more users?

    How can I boost the revenue? How can I save the cost? How can I keep my customers alive? (SBQI &NCFEEJOH PO6TFS "DRVJTJUJPO 6QMJGUPO"E QFSGPSNBODF #PPTUJOH "MHPSJUINPO $VTUPNFS -JGFUJNFWBMVF
  5. Marketing Align the User Journey Content & Experience Social &

    Relationship Commerce & Sales Management Data Advertising & Promotion • Talent Mgt. • Finance Mgt. • Quality Mgt. • CRM • Channels & Partners • Sales Automation & Intelligence • CDP/DMP • Business Intelligence • Data Governance • Customer Exp. • KOL • Community & Review • Content • Experience Mgt. • SEO • Marketing • Advertising • PR
  6. Apply graph embedding solutions to invite new service users. User

    Acquisition via Advertising:
  7. Across Multiple services Correlate users with behaviors of other services

    .VTJD 4IPQQJOH $PNJD ?
  8. Graph embedding can help Service knowledge graph provides abundant information

    1 2 https://arxiv.org/pdf/1710.10903.pdf https://arxiv.org/pdf/1706.02216.pdf 1 2 Algorithm choice depends on data traits › Semantic matching based: ComplEX 
 › GraphSage algorithm 
 › Graph Attention network with deepwalk
  9. New acquired users Case study on LINE POINTS wall :

    how to acquire new users? -*/&10*/54XBMMVTFST OPO-*/&10*/54XBMMVTFST 0 ffi DJBM "DDPVOU GPMMPXJOH TUBUVT ,OPXMFEHFHSBQI NPEFMJOH $MVTUFSJOH ,OPXMFEHF HSBQINPEFMJOH 
 $MVTUFSJOH OPO-*/&10*/54 XBMMVTFST NPTUMJLFMZUP KPJO-*/&SFXBSE Compose triples Compose triples Top k similar users
  10. - Flexible and complex decision making - Embeddings show clustering

    effect w.r.t campaign New acquired users Case study on LINE POINTS wall : effectiveness and fully data orientation User followed OA category target size in 202106 target size in 202107 KG target size (tunable) Type A 2.2M 3.1M 500k Type B 10.4M 10.2M Type C 3.3M 4.1M
  11. New acquired users Case study on LINE Reward : Steady

    growth of new users Broadcasted user count New user count
  12. Takeaway User Acquisition on LINE Reward 1. Manual Decision Making

    of Audience Filtering Is Exhausting, but Martech Is More Flexible 2. Graph-Based Model Could Handle a Large Scale of Data With Reasonable Logics.
  13. Apply uplift model to target the persuadable Advertising: How to

    Maximize Campaign Profit?
  14. Boost the Profits with Incentives $PVOUPGDBNQBJHO    

     .POUIPG:FBS             3FDFJWF 0"NFTTBHF 1VSDIBTF JODBNQBJHOQFSJPE &BSO -*/&$0*/-*/&10*/54
  15. User Segments by Uplift High Low Low High Buy if

    treated Buy if not treated Sure Thing Sleeping Dog Lost Cause The Persuadable
  16. User Segments by Uplift EFMJWFSDPTU JODFOUJWFDPTU High Low Low High

    Buy if treated Buy if not treated Sure Thing Sleeping Dog Lost Cause The Persuadable
  17. User Segments by Uplift EFMJWFSZDPTU SFWFOVF High Low Low High

    Buy if treated Buy if not treated Sure Thing Sleeping Dog Lost Cause The Persuadable
  18. User Segments by Uplift EFMJWFSZDPTU High Low High Low Buy

    if treated Buy if not treated Sure Thing Sleeping Dog Lost Cause The Persuadable
  19. User Segments by Uplift  EFMJWFSZDPTU  *ODFOUJWFDPTU  SFWFOVF

    High Low Low High Buy if treated Buy if not treated Sure Thing Sleeping Dog Lost Cause The Persuadable
  20. User Segments by Uplift QPTJUJWFVQMJGU OFHBUJWFVQMJGU OPVQMJGU OPVQMJGU High Low

    Low High Buy if treated Buy if not treated Sure Thing Sleeping Dog Lost Cause The Persuadable Uplift M odel
  21. User Segments by Uplift High Low Low High Buy if

    treated Buy if not treated Sure Thing Sleeping Dog Lost Cause The Persuadable 3FTQPOTJWF.PEFM Uplift M odel
  22. A/B Testing to Collect Lookalike User Groups Treatment Group Control

    Group
  23. User 360 Degree Features Campaign Period Observation Period, 30 days

    Purchase or not Purchase or not Browsing behavior Demographics Purchase history OA engagement Ad event Sticker collections Treatment Group Control Group
  24. User 360 Degree Features Campaign Period Observation Period, 30 days

    Purchase or not Purchase or not Browsing behavior Demographics Purchase history OA engagement Ad event Sticker collections Treatment Group Control Group If the campaign effect is significant  If the sure thing exist 
  25. Meta-learners, S-Learner Campaign Period Observation Period, 30 days Browsing behavior

    Demographics Purchase history OA engagement Ad event Sticker collections Purchase probabilities Treatment Group Control Group $MBTTJ fi FS treatment = 1 treatment = 0 Purchase or not ≈  *OUVJUJWFBOEFBTZUPUSBJO  -PXWBSJBODFBOE fl FYJCMF
  26. Estimate Individual Treatment Effects Campaign Period Observation Period, 30 days

    Browsing behavior Demographics Purchase history OA engagement Ad event Sticker collections $MBTTJ fi FS Purchase probability Purchase probability treatment = 1 treatment = 0 candidate
  27. Estimate Individual Treatment Effects Campaign Period Observation Period, 30 days

    Browsing behavior Demographics Purchase history OA engagement Ad event Sticker collections $MBTTJ fi FS treatment = 1 treatment = 0 Uplift Purchase probability candidate
  28. Uplifts by Deciles 6QMJGU      

             The Persuadable Sure Thing/Lost Cause Sleeping Dog
  29. Facilitate LINE STICKER to Optimize Campaign Performances ROI OAP Profit

    -60% +10%
  30. Lesson Learnt Consider the user segmentation with uplift A well-designed

    experiment is the must
  31. Actionable and Explainable User Segmentation from RFM to CLV Customer

    Lifetime Value: How to identify valuable customers?
  32. Know Your Customers ) 5 0 Historical Unknown Time Now

    From Buying Profile
  33. Measure Your Customer Lifetime Value From RFM to CLV Past

    Now Future Measure Former Value Predict Future Value RFM CLV
  34. Measure Your Customer Lifetime Value From RFM to CLV Past

    Now Future RFM CLV Measure Former Value Predict
  35. Measure Former Value - RFM Modeling RFM Scores Historical data

    R F M Segments
  36. Measure Former Value - RFM Modeling R F M RFM

    Segments Historical data Clustering
  37. RFM Modeling - Clustering R F

  38. RFM Modeling - Clustering R F

  39. Measure Former Value - RFM Modeling R F M Historical

    data Clustering
  40. Measure Your Customer Lifetime Value From RFM to CLV Past

    Now Future RFM CLV Measure Former Value Predict Future Value
  41. Predict Future Value - CLV Modeling Demographic Browsing history Buying

    pro fi le OA interaction Feature Engineering Modeling Tags CLV High ↑ CLV Median CLV Low ↓ { > 50% High valued customers <= 50% Medium valued customers 0, Churned customers 450 days 180 days Now Cascade model
  42. Predict Future Value - CLV Modeling Feature Engineering Modeling Tags

    CLV High ↑ CLV Median CLV Low ↓ { > 50% High valued customers <= 50% Medium valued customers 0, Churned customers 450 days 180 days Now User Embedding Demographic Browsing history Buying pro f i
  43. Predict Future Value - CLV Modeling User Embedding Time of

    View Content Views from customers AD 1 AD n Product 1 Article 1 Article n Product n … … … Context window
  44. Actionable and Explainable User Segmentation R↑F↑M↑ CLV Low Retain !

  45. Actionable and Explainable User Segmentation R↑F↑M↓ CLV High Upselling !

  46. Facilitate LINE TRAVEL TW

  47. Acquire New User Maximize Campaign Profit Manage Customer Relationship Graph

    Embed. Uplift CLV MarTech Solution Pipelines
  48. Scalability of MarTech How to build General MarTech solution for

    each service?
  49. Pain of Reusability & Scalability 4PMVUJPO1JQFMJOF" 4PMVUJPO1JQFMJOF$ 4PMVUJPO1JQFMJOF% 4PMVUJPO1JQFMJOF# Sources

    Pipelines Service A Service C Service B Service D
  50. Centralization and Automation Data Source %BUB 8BSFIPVTF Data Integration 'FBUVSF4UPSF

    Application "DRVJSFOFXVTFS "%5BSHFUJOH #PPTUSFWFOVF *NQSPWF3FUFOUJPO … ,(4PMVUJPO1JQFMJOF $-74PMVUJPO1JQFMJOF 6QMJGU 
 4PMVUJPO1JQFMJOF AI Solution 6TFSTFHNFOUBUJPO 
 4PMVUJPO1JQFMJOF … LINE Family Services Demographic Log
  51. Applied MarTech Services LINE Family Services Benefit Content & Experience

    Social & Relationship Commerce & Sales Management Data Advertising & Promotion
  52. AI Marketing Tech Can Help

  53. Thank you