ログから生まれる施策 / actions born from logs

WEB DIRECTORS SESSION Vol.2
http://peatix.com/event/219845

Transcript

1. ϖύϘͷϩάz׆༻zج൫ʮ#JHGPPUʯ
   ࡒ௡େՆ(.01FQBCP *OD
   8%*3&$50344&44*0/7PM
   ϩά͔Βੜ·ΕΔࢪࡦ
   

   View Slide

2. σʔλαΠΤϯςΟετσΟϨΫλʔ
   ࡒ௡େՆ![BJNZ
   NJOOFࣄۀ෦
   

   View Slide

3. ϋϯυϝΠυϚʔέοτNJOOF
   IUUQTNJOOFDPN
   

   View Slide

4. ໨࣍
   w8FCαʔϏεͷߦಈϩά
   wϖύϘͷϩάz׆༻zج൫ʮ#JHGPPUʯ
   wࢪࡦ΁ͷ׆༻ྫ
   

   View Slide

5. 8FCαʔϏεͷߦಈϩά
   

   View Slide

6. 8FCαʔϏεͷߦಈϩά
   wϢʔβʔ͕αʔϏεΛར༻ͨ͠ࡍͷཤྺ
   wʮ͍ͭʯʮ୭͕ʯʮԿΛʯߦͬͨͷ͔
   w݁Ռ͚ͩͰ͸ͳͦ͘͜ʹࢸΔաఔ΍ɺ݁Ռ్தͰఘΊͯ͠·ͬͨϢʔ
   βʔͷߦಈ΋෼͔Δ
   

   View Slide

7. ߦಈϩάΛ࢖ͬͯ
   ΑΓྑ͍αʔϏε΁!
   

   View Slide

8. ׆༻·Ͱͷஈ֊
   wऩूߦಈϩά͕ग़ྗ͞ΕɺऔΓ·ͱΊΒΕ͍ͯΔঢ়ଶ
   w෼ੳऔΓ·ͱΊͨߦಈϩάΛࢹ֮Խɺ෼ੳͰ͖Δঢ়ଶ
   w׆༻෼ੳͨ͠ߦಈϩάΛ΋ͱʹܧଓతͳαʔϏεվળ͕ߦ͍͑ͯΔঢ়ଶ
   

   View Slide

9. #JHGPPU
   IUUQTJDPOTDPN
   

   View Slide

10. #JHGPPU
    wϖύϘͷϩάz׆༻zج൫
    w෯޿͍δϟϯϧͷ8FCαʔϏε
    wϋϯυϝΠυϚʔέοτNJOOF
    wωοτγϣοϓ࡞੒Χϥʔϛʔγϣοϓ
    wϩϦϙοϓʂϨϯλϧαʔόʔ
    wϩάऩू͔Β׆༻·Ͱͷ֤ஈ֊ʹ͓͍ͯ൚༻తʹར༻Ͱ͖Δશࣾج൫
    

    View Slide

11. #JHGPPU
    rack-bigfoot
    Service
    Request
    Activity
    log
    Services
    DB
    Attribute
    Big Cube
    Cube
    https://speakerdeck.com/monochromegane/pepabo-log-infrastructure-bigfoot
    Bandit algorithm/
    Recommendation
    Re-targeting
    Feedback
    Name identification
    BI/Visualize
    

    View Slide

12. ऩू
    

    View Slide

13. ໊دͤ
    wαʔϏεͷΞΧ΢ϯτͱΫϥΠΞϯτʢ୺຤ɾϒϥ΢βʣΛඥ͚ͮ
    wैདྷ͸ΫϥΠΞϯτ͝ͱͷܭଌ
    wະϩάΠϯঢ়ଶͷΞΧ΢ϯτ΋ͦͷޙϩάΠϯͨ͠λΠϛϯάͰաڈʹ
    Ḫͬͯඥ͚ͮΒΕΔ
    wαʔϏεΛލ͍ͩඥ͚ͮ΋Մೳ
    

    View Slide

14. ෼ੳ
    

    View Slide

15. #JH$VCFͱ$VCF
    w#JHGPPU্ͷશͯͷϩάΛ#JH$VCFʹू໿
    w෼ੳͷ੾ΓޱʢσΟϝϯγϣϯͱϝδϟʔʣ͕ܾ·ͬͨΒ$VCFʹूܭ
    wྫʣ࣌ؒ͝ͱ஫จֹۚ
    wߴ଎ͳࢀর͕Մೳ
    Activity Big Cube Cube
    HiveQL SQL
    Dashboard
    Ad-hoc query
    Analyst Managers,
    Product owners,
    Promotion groups
    

    View Slide

16. ࢹ֮Խ
    w5BCMFBV
    wIUUQXXXUBCMFBVDPN
    wΞυϗοΫͳ෼ੳ΍ෳࡶͳσʔλ݁߹ʹ
    w3FEBTI
    wIUUQTSFEBTIJP
    w୭΋͕ࢀরͰ͖ΔμογϡϘʔυʹ
    

    View Slide

17. ύε෼ੳ
    wίϯόʔδϣϯʹؔ͢Δ෼ੳͷ͏ͪɺܦ࿏ʢύεʣʹओ؟Λஔ͍ͨ΋ͷ
    wύεͷ௕࣌ؒ͞ྨܕ
    wΞτϦϏϡʔγϣϯϞσϧϖʔδͷग़ݱҐஔʹΑΓॏΈΛม͑Δ
    referrer landing last
    cv
    ???
    Point Analytics
    referrer landing last
    Path Analytics
    cv
    

    View Slide

18. ࢪࡦ΁ͷ׆༻
    

    View Slide

19. ཭୤Ϣʔβʔ΁ͷϦςϯγϣϯ
    wߦಈϩά͔Β཭୤ϢʔβʔΛநग़
    wΧʔτʹ౸ୡ͕ͨ͠ങΘͳ͔ͬͨϢʔβʔʢ͍ΘΏΔΧʔτམͪʣ
    w୹ظؒʹಉ͡࡞඼ΛԿ౓΋ݟ͍ͯΔϢʔβʔ
    wಛఆͷ৚݅ʹ߹க͢Ε͹ϓογϡ௨஌΍ϝʔϧͰϦςϯγϣϯ
    wϨγʔτϝʔϧͱಉ౳ͷ։෧཰
    HiveQL
    Re-targeting
    

    View Slide

20. ޿ࠂ࿈ܞ
    wߦಈϩά͔ΒϢʔβʔΛηάϝϯτ
    wϢʔβʔʹؔ࿈ੑͷߴ͍޿ࠂΛදࣔ
    wطʹϦʔνͨ͠Ϣʔβʔʹ͸޿ࠂΛදࣔ͠ͳ͍ʢσϦλʔήςΟϯάʣ
    

    View Slide

21. όϯσΟοτΞϧΰϦζϜ
    w࡞඼ݕࡧը໘͔ΒͷརӹΛ࠷େԽ͍ͨ͠
    w$53ͷҟͳΔύλʔϯͷ࡞඼ݕࡧΞϧΰϦζϜ
    wച্ʹର͢Δൺॏ͕ଟ͍ҝʹγεςϜมߋͷϦεΫ͕ߴ͍
    w࠷ྑͷύλʔϯΛ࢖͍ͳ͕ΒΑΓྑ͍ύλʔϯΛಈతʹ୳͢
    

    View Slide

22. όϯσΟοτΞϧΰϦζϜ
    w&QTJMPO(SFFEZ"MHPSJUIN
    w֬཰ Џ
    Ͱͦͷ࣌఺ͷ࠷΋ظ଴஋͕ߴ͍ύλʔϯΛ࢖༻ʢ׆༻ʣ
    w֬཰ЏͰϥϯμϜʹબ୒ͨ͠ύλʔϯΛ࢖༻ʢ୳ࡧʣ
    Activity
    Epsilon-Greedy algorithm
    User
    1-ε: exploitation
    ε/pattern: exploration
    Click / Not click
    Import
    

    View Slide

23. όϯσΟοτΞϧΰϦζϜ
    wτϨʔυΦϑͷղܾ
    w׆༻͹͔Γ͍ͯ͠Δͱݱࡏͷظ଴஋ΑΓ΋ྑ͍࿹Λݟ͚ͭΒΕͳ͍
    w୳ࡧ͹͔Γ͍ͯ͠Δͱظ଴஋͕ߴ͍࿹͕ར༻͞Εͳ͍
    

    View Slide

24. Ϩίϝϯσʔγϣϯ
    wڠௐϑΟϧλϦϯάʹΑΔʮ͋ͳͨʹ͓͢͢Ίͷ࡞Ոʯ
    !NPOPDISPNFHBOFʮNJOOFNFFUT)JWFNBMMʯ
    IUUQTTQFBLFSEFDLDPNNPOPDISPNFHBOFQFQBCPNJOOFNBUSJYGBDUPSJ[BUJPOJOIJWFNBMM
    

    View Slide

25. ϩάz׆༻zج൫
    

    View Slide

26. ϩάz׆༻zج൫
    wᶃਓ͕ؒ෼ੳ݁ՌΛݟͯߦ͏੩తͳʮ׆༻ʯ
    wࢪࡦͷਫ਼౓޲্
    wᶄࣗಈԽ͞ΕͨϑΟʔυόοΫʹΑΔಈతͳʮ׆༻ʯ
    wͳΊΒ͔ͳγεςϜ
    

    View Slide