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

Conserved capacity of app usage - Data Science Lecture

Marco De Nadai
November 19, 2020

Conserved capacity of app usage - Data Science Lecture

Marco De Nadai

November 19, 2020
Tweet

More Decks by Marco De Nadai

Other Decks in Education

Transcript

  1. Strategies and limitations in app usage
    and human mobility
    Marco De Nadai (http://www.marcodena.it)

    View full-size slide

  2. 3
    Mobility, Social interactions are conserved

    View full-size slide

  3. 4
    Mobility, Social interactions are conserved

    View full-size slide

  4. 5
    What about apps?

    View full-size slide

  5. 6
    Photo: Liz Hafalia, The Chronicle + NBC News

    View full-size slide

  6. 7
    THERE ARE 2.1M APPS IN THE
    GOOGLE PLAY STORE
    App usage statistics by Statista.com & hubspot.net

    View full-size slide

  7. 8
    PEOPLE SPEND AN INCREASINGLY
    AMOUNT OF TIME ON THE PHONE […]
    REACHING 3 HOURS PER DAY
    THERE ARE 2.1M APPS IN THE
    GOOGLE PLAY STORE
    App usage statistics by Statista.com & hubspot.net

    View full-size slide

  8. 9
    PEOPLE SPEND AN INCREASINGLY
    AMOUNT OF TIME ON THE PHONE […]
    REACHING 3 HOURS PER DAY
    THERE ARE 2.1M APPS IN THE
    GOOGLE PLAY STORE
    App usage statistics by Statista.com & hubspot.net
    USER ENGAGMENT INCREASED BY 23%

    View full-size slide

  9. 10
    But I always use the same apps!

    View full-size slide

  10. 11
    ALWAYS USED
    But I always use the same apps!

    View full-size slide

  11. 12
    ALWAYS USED
    But I always use the same apps!

    View full-size slide

  12. 13
    ALWAYS USED
    But I always use the same apps!

    View full-size slide

  13. 14
    ALWAYS USED
    MEH... SOMETIMES
    But I always use the same apps!

    View full-size slide

  14. 15
    ALWAYS USED
    MEH... SOMETIMES
    But I always use the same apps!

    View full-size slide

  15. 16
    ALWAYS USED
    MEH... SOMETIMES
    WAS THIS APP EVEN
    HERE?
    But I always use the same apps!

    View full-size slide

  16. 17
    ALWAYS USED
    MEH... SOMETIMES
    WAS THIS APP EVEN
    HERE?
    • APP USAGE CHANGES
    • APPS CHANGE
    But I always use the same apps!

    View full-size slide

  17. 18
    ALWAYS USED
    MEH... SOMETIMES
    WAS THIS APP EVEN
    HERE?
    • APP USAGE CHANGES
    • APPS CHANGE
    But I always use the same apps!

    View full-size slide

  18. 19
    Monotony
    vs
    Growth of app usage

    View full-size slide

  19. 20
    •Screen time of 90K people
    •Over 8 months
    •69K different apps
    •12M of people’s hours spent on apps
    Data

    View full-size slide

  20. 21
    Data (pseudo-anonymized)

    View full-size slide

  21. 22
    Data (pseudo-anonymized)
    •Screen time
    •No network-based approaches
    •Unique dataset of app usage

    View full-size slide

  22. 23
    • !
    ~ ( + "
    )#$#!
    "
    • = 1.19 ± 0.01
    Characterizing App usage

    View full-size slide

  23. 24
    • !
    ~ ( + "
    )#$#!
    "
    • = 1.19 ± 0.01
    Characterizing App usage
    • !
    ~ ()#$
    • = 1.27 ± 0.01

    View full-size slide

  24. 25
    • !
    ~ ( + "
    )#$#!
    "
    • = 1.19 ± 0.01
    Characterizing App usage
    • !
    ~ ()#$
    • = 1.27 ± 0.01
    • ~ %
    • γ = 0.41

    View full-size slide

  25. 26
    • !
    ~ ( + "
    )#$#!
    "
    • = 1.19 ± 0.01
    Characterizing App usage
    • !
    ~ ()#$
    • = 1.27 ± 0.01
    • ~ %
    • γ = 0.41
    • PEOPLE TIME IS FOCUSED ON FEW APPS
    • PEOPLE KEEP EXPLORING NEW APPS

    View full-size slide

  26. 27
    The familiar apps over time: the App space
    THE APP SPACE (t)

    View full-size slide

  27. 28
    • Apps used at least twice and for at least 10 min per week
    • 20 weeks long sliding windows (1 week)
    • How does the App space evolve over time?
    The familiar apps over time: the App space
    THE APP SPACE (t)
    THE APP SPACE (t-1) THE APP SPACE (t+1)
    1 WEEK 1 WEEK

    View full-size slide

  28. App space over time (adopted and dropped apps)
    THE APP SPACE (t)

    View full-size slide

  29. App space over time (adopted and dropped apps)
    !
    THE APP SPACE (t)

    View full-size slide

  30. App space over time (adopted and dropped apps)
    !
    ! = (!) = 4
    THE APP SPACE (t)

    View full-size slide

  31. App space over time (adopted and dropped apps)
    "
    !
    ! = (!) = 4
    " = (") = 5
    THE APP SPACE (t)

    View full-size slide

  32. App space over time (adopted and dropped apps)
    #
    "
    !
    ! = (!) = 4
    " = (") = 5
    # = (#) = 4
    THE APP SPACE (t)

    View full-size slide

  33. App space over time (adopted and dropped apps)
    !$"
    =
    !$"
    =
    "$#
    =
    "$# =
    #
    "
    !
    ! = (!) = 4
    " = (") = 5
    # = (#) = 4
    THE APP SPACE (t)

    View full-size slide

  34. 35
    • Gain: 3
    = 3
    − 3
    ()
    • 97.5% of people exhibit a conserved capacity ( 4#
    5$#
    ≤ 1)
    Adoped and dropped apps

    View full-size slide

  35. What is the conserved capacity?
    On average:
    •Everytime a new app is adopted, an older app is discarded
    THE APP SPACE (t)

    View full-size slide

  36. What is the conserved capacity?
    On average:
    •Everytime a new app is adopted, an older app is discarded
    THE APP SPACE (t)

    View full-size slide

  37. What is the conserved capacity?
    On average:
    •Everytime a new app is adopted, an older app is discarded
    THE APP SPACE (t)

    View full-size slide

  38. What is the conserved capacity?
    On average:
    •Everytime a new app is adopted, an older app is discarded
    THE APP SPACE (t)

    View full-size slide

  39. 40
    •Local shuffle: app usage order of a user
    •If the random permutated user has the same capacity => Time
    •Two-sample Kolmogorov–Smirnov test
    •Set p-value threshold (0.05 or 0.001?)
    - Different (p-value < threshold)
    - Don’t know (p-value > threshold)
    Is capacity a time-constraints consequence?
    !
    "
    #
    $
    USER A #
    $
    !
    "
    https://en.wikipedia.org/wiki/Kolmogorov%E2%80%93Smirnov_test

    View full-size slide

  40. 41
    •Local shuffle: app usage order of a user
    No time consequence (KS: 0.55 p-value < 0.001).
    •Global: app usage among users
    (KS: 0.98 p-value < 0.001). It’s an individual characteristics.
    Is capacity a time-constraints consequence?
    !
    "
    #
    $
    USER A #
    $
    !
    "
    !
    "
    #
    $
    USER A !
    "
    #
    $
    !
    "
    #
    $
    USER B !
    "
    #
    $

    View full-size slide

  41. 42
    The app strategy
    • EXPLORERS (% ≫ )
    • KEEPERS (%
    ≪ )
    • Explorers adopt 1 new app every 28 weeks
    • Keepers always keep using the same apps
    % = % / %

    View full-size slide

  42. 43
    The app strategy
    • EXPLORERS (% ≫ )
    • KEEPERS (%
    ≪ )
    • Explorers adopt 1 new app every 28 weeks
    • Keepers always keep using the same apps
    % = % / %

    View full-size slide

  43. 44
    The app strategy
    • EXPLORERS (% ≫ )
    • KEEPERS (%
    ≪ )
    • Explorers adopt 1 new app every 28 weeks
    • Keepers always keep using the same apps
    % = % / %

    View full-size slide

  44. 45
    What about age?

    View full-size slide

  45. 46
    What about age?

    View full-size slide

  46. 47
    What about age?
    16
    What about age?

    View full-size slide

  47. 48
    People are focused on few apps, but keep exploring.
    Individuals exhibit a conserved app capacity
    •Capacity is an individual behaviour
    •Capacity varies with age
    •We can define app explorers and keepers
    •AND MORE!
    So?

    View full-size slide

  48. 52
    Apps, Mobility, Social interactions are conserved
    DUNBAR’S NUMBER OF APPS?

    View full-size slide

  49. 53
    Apps that stay the most
    ~10% of apps are always kept
    ~17.5% are continuosly changed

    View full-size slide

  50. 54
    Strategy vs age

    View full-size slide

  51. 55
    EXERCISEEEE
    www.menti.com – Code: 2145921

    View full-size slide

  52. 56
    EXERCISEEEE!
    shorturl.at/nwSZ8

    View full-size slide

  53. 65
    What did I learn?

    View full-size slide

  54. 66
    Individuals exhibit a conserved app capacity
    •To explore dataaa
    •Understand complex queries
    •A bit of spark
    What did I learn?

    View full-size slide

  55. 67
    Individuals exhibit a conserved app capacity
    •To explore dataaa
    •Understand complex queries
    •A bit of spark
    What did I learn?

    View full-size slide

  56. 68
    Individuals exhibit a conserved app capacity
    •To explore dataaa
    •Understand complex queries
    •A bit of spark
    What did I learn?

    View full-size slide

  57. Thank you!
    @denadai2

    View full-size slide