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

Analyze this… Analytics Strategies for iOS and Rails

Analyze this… Analytics Strategies for iOS and Rails

You've just build a killer app with a killer website. Downloads are skyrocketing and you are getting lots of good feedback from users. Then someone asks you for data. Where are people using the app? How many times have people used a certain feature? Can you make predictions about user interactions based on past usage? In this session, let's take a look at what goes into mining mobile data and analyzing the insights. We'll discuss different approaches to gathering, storing, and querying data, which metrics and KPIs to focus on, and how to interpret and use information in order to determine the real drivers behind user behavior.

Lance Gleason

April 13, 2013
Tweet

More Decks by Lance Gleason

Other Decks in Technology

Transcript

  1. Twitter @lgleasain
    Github lgleasain
    www.lancegleason.com
    www.polyglotprogrammincinc.com
    [email protected]
    1
    Sunday, April 14, 13

    View full-size slide

  2. Introductions
    2
    Sunday, April 14, 13

    View full-size slide

  3. 3
    Sunday, April 14, 13

    View full-size slide

  4. 4
    Sunday, April 14, 13

    View full-size slide

  5. 5
    Sunday, April 14, 13

    View full-size slide

  6. http://www.polyglotprogramminginc.com/purr-
    programming-2-0/
    6
    Sunday, April 14, 13

    View full-size slide

  7. Data Science
    7
    Sunday, April 14, 13

    View full-size slide

  8. 8
    Sunday, April 14, 13

    View full-size slide

  9. 8
    Sunday, April 14, 13

    View full-size slide

  10. 8
    Sunday, April 14, 13

    View full-size slide

  11. 9
    Sunday, April 14, 13

    View full-size slide

  12. 10
    Sunday, April 14, 13

    View full-size slide

  13. Analytics
    11
    Sunday, April 14, 13

    View full-size slide

  14. 12
    Sunday, April 14, 13

    View full-size slide

  15. 13
    Sunday, April 14, 13

    View full-size slide

  16. http://www.torlaune.de/euro-2012/spieler-relationen/
    http://www.nytimes.com/interactive/2012/05/17/business/
    dealbook/how-the-facebook-offering-compares.html?_r=0
    http://www.nytimes.com/interactive/2012/08/24/us/drought-
    crops.html
    14
    Sunday, April 14, 13

    View full-size slide

  17. ?
    15
    Sunday, April 14, 13

    View full-size slide

  18. 16
    Sunday, April 14, 13

    View full-size slide

  19. Which Customers
    Bought the Most?
    17
    Sunday, April 14, 13

    View full-size slide

  20. Which Will Buy
    the Most?
    18
    Sunday, April 14, 13

    View full-size slide

  21. http://www.kaggle.com/c/titanic-gettingStarted
    19
    Sunday, April 14, 13

    View full-size slide

  22. 20
    Sunday, April 14, 13

    View full-size slide

  23. Coupon App
    21
    Sunday, April 14, 13

    View full-size slide

  24. 22
    Sunday, April 14, 13

    View full-size slide

  25. •Users can search for deals via proximity
    to location
    22
    Sunday, April 14, 13

    View full-size slide

  26. •Users can search for deals via proximity
    to location
    •Customers can save coupons
    22
    Sunday, April 14, 13

    View full-size slide

  27. •Users can search for deals via proximity
    to location
    •Customers can save coupons
    •When a coupon is used it is no longer
    available for use
    22
    Sunday, April 14, 13

    View full-size slide

  28. •Users can search for deals via proximity
    to location
    •Customers can save coupons
    •When a coupon is used it is no longer
    available for use
    •Customers can search via categories
    22
    Sunday, April 14, 13

    View full-size slide

  29. •Users can search for deals via proximity
    to location
    •Customers can save coupons
    •When a coupon is used it is no longer
    available for use
    •Customers can search via categories
    •Advertisers pay for data and preferential
    ad pacement
    22
    Sunday, April 14, 13

    View full-size slide

  30. How many sales are being generated
    by this app?
    23
    Sunday, April 14, 13

    View full-size slide

  31. Am I getting more active users?
    24
    Sunday, April 14, 13

    View full-size slide

  32. How many users are saving coupons?
    25
    Sunday, April 14, 13

    View full-size slide

  33. How many users are saving coupons?
    •Which stores have the most saves.
    25
    Sunday, April 14, 13

    View full-size slide

  34. How many users are saving coupons?
    •Which stores have the most saves.
    •Do certain categories get more saves?
    25
    Sunday, April 14, 13

    View full-size slide

  35. How many are using coupons?
    26
    Sunday, April 14, 13

    View full-size slide

  36. How many are using coupons?
    •Which stores have the most uses?
    26
    Sunday, April 14, 13

    View full-size slide

  37. How many are using coupons?
    •Which stores have the most uses?
    •Do certain categories get used more?
    26
    Sunday, April 14, 13

    View full-size slide

  38. How many are using coupons?
    •Which stores have the most uses?
    •Do certain categories get used more?
    •Which physical store is the user in?
    26
    Sunday, April 14, 13

    View full-size slide

  39. Which stores are they at?
    27
    Sunday, April 14, 13

    View full-size slide

  40. Hidden Insights
    28
    Sunday, April 14, 13

    View full-size slide

  41. 29
    Sunday, April 14, 13

    View full-size slide

  42. 30
    Sunday, April 14, 13

    View full-size slide

  43. Appstore Data
    31
    Sunday, April 14, 13

    View full-size slide

  44. 32
    Sunday, April 14, 13

    View full-size slide

  45. 32
    Sunday, April 14, 13

    View full-size slide

  46. 32
    Sunday, April 14, 13

    View full-size slide

  47. 32
    Sunday, April 14, 13

    View full-size slide

  48. 32
    Sunday, April 14, 13

    View full-size slide

  49. 33
    Sunday, April 14, 13

    View full-size slide

  50. 34
    Sunday, April 14, 13

    View full-size slide

  51. 35
    Sunday, April 14, 13

    View full-size slide

  52. 36
    Sunday, April 14, 13

    View full-size slide

  53. 37
    Sunday, April 14, 13

    View full-size slide

  54. Logging
    (Papertrail/
    Loggly)
    37
    Sunday, April 14, 13

    View full-size slide

  55. Logging
    (Papertrail/
    Loggly)
    Amazon S3
    37
    Sunday, April 14, 13

    View full-size slide

  56. {"measure":"instance","instance":
    "stores","store_id":
    64696,"company_id":
    210,"store_name":"bebe",
    "controller":"api/v1/
    stores","action":"index"}
    38
    Sunday, April 14, 13

    View full-size slide

  57. 39
    Sunday, April 14, 13

    View full-size slide

  58. Amazon Elastic
    Map Reduce
    39
    Sunday, April 14, 13

    View full-size slide

  59. Amazon Elastic
    Map Reduce
    39
    Sunday, April 14, 13

    View full-size slide

  60. DynamoDB
    40
    Sunday, April 14, 13

    View full-size slide

  61. CREATE EXTERNAL TABLE events_1 (
    id bigint, received_at string, generated_at string, source_id
    bigint, source_name string, source_ip string, facility string,
    severity string, program string, message string
    )
    PARTITIONED BY (
    dt string
    )
    ROW FORMAT DELIMITED FIELDS TERMINATED BY '\t'
    STORED AS TEXTFILE
    LOCATION 's3://mybucket/papertrail/logs/production';
    41
    Sunday, April 14, 13

    View full-size slide

  62. ALTER TABLE events_1 RECOVER PARTITIONS;
    42
    Sunday, April 14, 13

    View full-size slide

  63. CREATE EXTERNAL TABLE promotions_1 (id string,
    received_at string, source_id string, source_ip string,
    source_name string,measure string, instance string,
    promotion_id string, company_id string, controller string,
    action string)
    stored by
    'org.apache.hadoop.hive.dynamodb.DynamoDBStorageHan
    dler'
    TBLPROPERTIES ("dynamodb.table.name" =
    "sh_promotions_latest",
    "dynamodb.column.mapping" =
    "id:id,received_at:received_at,source_id:source_id,source_i
    p:source_ip,source_name:source_name,measure:measure,i
    nstance:instance,promotion_id:promotion_id,company_id:c
    ompany_id,controller:controller,action:action");
    43
    Sunday, April 14, 13

    View full-size slide

  64. alter table promotions_1 recover partitions;
    44
    Sunday, April 14, 13

    View full-size slide

  65. insert overwrite table promotions_1 select id, received_at,
    source_id, source_ip, source_name,
    get_json_object(message, '$.measure') as measure,
    get_json_object(message, '$.instance') as instance,
    get_json_object(message, '$.promotion_id') as
    promotion_id, get_json_object(message, '$.company_id') as
    company_id, get_json_object(message, '$.controller') as
    controller, get_json_object(message, '$.action') as action
    from events_1 where message like '%"promotion"%' ;
    45
    Sunday, April 14, 13

    View full-size slide

  66. 46
    Sunday, April 14, 13

    View full-size slide

  67. 47
    Sunday, April 14, 13

    View full-size slide

  68. d3js.org
    48
    Sunday, April 14, 13

    View full-size slide

  69. 49
    Sunday, April 14, 13

    View full-size slide

  70. 49
    Sunday, April 14, 13

    View full-size slide

  71. 49
    Sunday, April 14, 13

    View full-size slide

  72. False Positives
    50
    Sunday, April 14, 13

    View full-size slide

  73. 51
    Sunday, April 14, 13

    View full-size slide

  74. Nearly ALL sick people have
    eaten PEAS (obviously then,
    the effects
    are cumulative).
    52
    Sunday, April 14, 13

    View full-size slide

  75. An estimated 99.9% of all
    people who die from cancer
    or heart attacks
    have eaten PEAS.
    53
    Sunday, April 14, 13

    View full-size slide

  76. Another 99.9% of people
    involved in auto accidents ate
    PEAS within
    60-days before the accident.
    54
    Sunday, April 14, 13

    View full-size slide

  77. Among people born in 1839
    who later dined on PEAS,
    there has been a
    100% mortality rate
    55
    Sunday, April 14, 13

    View full-size slide

  78. Peas Will Kill You
    56
    Sunday, April 14, 13

    View full-size slide

  79. We had 4000 app downloads
    this month. We are doing
    great....
    57
    Sunday, April 14, 13

    View full-size slide

  80. 58
    Sunday, April 14, 13

    View full-size slide

  81. Most people use the app once
    and then uninstall it.
    59
    Sunday, April 14, 13

    View full-size slide

  82. 60
    Sunday, April 14, 13

    View full-size slide

  83. My shopping app just saw a
    spike in weekly usage after I
    made UI changes.
    61
    Sunday, April 14, 13

    View full-size slide

  84. That UI change led to more
    users!
    62
    Sunday, April 14, 13

    View full-size slide

  85. 63
    Sunday, April 14, 13

    View full-size slide

  86. The change went live during
    the last week of November.
    64
    Sunday, April 14, 13

    View full-size slide

  87. 65
    Sunday, April 14, 13

    View full-size slide

  88. Be Wary of N of 1
    Experiments
    66
    Sunday, April 14, 13

    View full-size slide

  89. Segmentation
    67
    Sunday, April 14, 13

    View full-size slide

  90. Sparse Data
    68
    Sunday, April 14, 13

    View full-size slide

  91. To Get Statistically
    Meaningful Results you
    will need thousands of
    data points
    69
    Sunday, April 14, 13

    View full-size slide

  92. The Results
    Need to Pass
    the Smell Test
    70
    Sunday, April 14, 13

    View full-size slide

  93. 71
    Sunday, April 14, 13

    View full-size slide

  94. Collect Lots
    of Data Early
    72
    Sunday, April 14, 13

    View full-size slide

  95. Go For Low
    Hanging Fruit
    First
    73
    Sunday, April 14, 13

    View full-size slide

  96. Try to Gather
    Data Rich Data
    Points (whenever
    possible)
    74
    Sunday, April 14, 13

    View full-size slide

  97. 75
    Sunday, April 14, 13

    View full-size slide

  98. Twitter @lgleasain
    Github lgleasain
    www.lancegleason.com
    www.polyglotprogrammincinc.com
    [email protected]
    76
    Sunday, April 14, 13

    View full-size slide