{
  "id": 54765,
  "title": "Business Context of Click Fraud - Chinese Click Farm",
  "url": "/competitions/talkingdata-adtracking-fraud-detection/discussion/54765",
  "author_name": "",
  "post_date": "2018-04-17T14:46:06.898771200Z",
  "votes": 73,
  "comment_count": 18,
  "views": 0,
  "content": "<p>Hello all,   I am supporting students from <a href=\"https://diveintocode.jp/ai_curriculum\">DiveIntoCode</a> to learn practical (and impractical) ML skills via Kaggle.  so it is always nice to start tackling a machine learning problem by understanding its business context. Furthermore, seems no one has gone to explore the business context of click fraud here, and I have come across this nice article about chinese click farm.</p>\n\n<p>So I hope by sharing this, you can get a better understanding of how these fake clicks are being generated, and who are doing it.  See below for a few interesting images:</p>\n\n<p><img src=\"https://snag.gy/oiyHMS.jpg\" alt=\"enter image description here\"></p>\n\n<p><img src=\"https://snag.gy/TYfjq5.jpg\" alt=\"enter image description here\"></p>\n\n<p>The full link is here: <a href=\"https://kotaku.com/inside-chinese-click-farms-1795287821\">https://kotaku.com/inside-chinese-click-farms-1795287821</a></p>",
  "messages": [
    {
      "id": "315752",
      "postDate": "04/17/2018 14:46:06",
      "content": "<p>Hello all,   I am supporting students from <a href=\"https://diveintocode.jp/ai_curriculum\">DiveIntoCode</a> to learn practical (and impractical) ML skills via Kaggle.  so it is always nice to start tackling a machine learning problem by understanding its business context. Furthermore, seems no one has gone to explore the business context of click fraud here, and I have come across this nice article about chinese click farm.</p>\n\n<p>So I hope by sharing this, you can get a better understanding of how these fake clicks are being generated, and who are doing it.  See below for a few interesting images:</p>\n\n<p><img src=\"https://snag.gy/oiyHMS.jpg\" alt=\"enter image description here\"></p>\n\n<p><img src=\"https://snag.gy/TYfjq5.jpg\" alt=\"enter image description here\"></p>\n\n<p>The full link is here: <a href=\"https://kotaku.com/inside-chinese-click-farms-1795287821\">https://kotaku.com/inside-chinese-click-farms-1795287821</a></p>",
      "rawMarkdown": "Hello all,   I am supporting students from [DiveIntoCode][1] to learn practical (and impractical) ML skills via Kaggle.  so it is always nice to start tackling a machine learning problem by understanding its business context. Furthermore, seems no one has gone to explore the business context of click fraud here, and I have come across this nice article about chinese click farm.\n\nSo I hope by sharing this, you can get a better understanding of how these fake clicks are being generated, and who are doing it.  See below for a few interesting images:\n\n![enter image description here][2]\n\n![enter image description here][3]\n\nThe full link is here: https://kotaku.com/inside-chinese-click-farms-1795287821\n\n\n  [1]: https://diveintocode.jp/ai_curriculum\n  [2]: https://snag.gy/oiyHMS.jpg\n  [3]: https://snag.gy/TYfjq5.jpg",
      "votes": null
    },
    {
      "id": "315797",
      "postDate": "04/17/2018 16:01:04",
      "content": "<p>It's really crazy what people are willing to do. There was a story of a group of guys carrying around 400,000 unregistered sim cards a while back. It really sucks if you're trying to break out as app developer to compete against automated clicks. <a href=\"https://www.thestar.com.my/tech/tech-news/2017/06/13/thai-police-bust-400000-sim-card-chinese-click-farm/\">https://www.thestar.com.my/tech/tech-news/2017/06/13/thai-police-bust-400000-sim-card-chinese-click-farm/</a></p>",
      "rawMarkdown": "It's really crazy what people are willing to do. There was a story of a group of guys carrying around 400,000 unregistered sim cards a while back. It really sucks if you're trying to break out as app developer to compete against automated clicks. https://www.thestar.com.my/tech/tech-news/2017/06/13/thai-police-bust-400000-sim-card-chinese-click-farm/",
      "votes": null
    },
    {
      "id": "315804",
      "postDate": "04/17/2018 16:12:09",
      "content": "<p>Wow !..This is beyond crazy..:p</p>\n\n<p>Thanks Yifan for the sharing :)</p>",
      "rawMarkdown": "Wow !..This is beyond crazy..:p\n\nThanks Yifan for the sharing :)",
      "votes": null
    },
    {
      "id": "315829",
      "postDate": "04/17/2018 16:34:57",
      "content": "<p>you are welcomed!  It really opened my eyes as well, even though I have heard about click farms before, seeing the images really hit home what kind of massive operation it is. </p>\n\n<p>I have more to share, but they are in Chinese so will take time to translate :)\nI guess I won't be seriously competing in this one, so getting a few discussion medals will do ^_^</p>",
      "rawMarkdown": "you are welcomed!  It really opened my eyes as well, even though I have heard about click farms before, seeing the images really hit home what kind of massive operation it is. \n\nI have more to share, but they are in Chinese so will take time to translate :)\nI guess I won't be seriously competing in this one, so getting a few discussion medals will do ^_^",
      "votes": null
    },
    {
      "id": "315830",
      "postDate": "04/17/2018 16:35:49",
      "content": "<p>The article states that click farms can be used for app rating manipulation by downloading and installing apps. In this competition we use the fact of app downloads as a sign of real user. This means we have filtered/pre-processed data, otherwise we would see much more app downloads.</p>",
      "rawMarkdown": "The article states that click farms can be used for app rating manipulation by downloading and installing apps. In this competition we use the fact of app downloads as a sign of real user. This means we have filtered/pre-processed data, otherwise we would see much more app downloads.",
      "votes": null
    },
    {
      "id": "316045",
      "postDate": "04/18/2018 04:34:23",
      "content": "<p>For a richer depiction, see a <a href=\"https://www.youtube.com/watch?v=Y-W0CBOGnnI\">clip</a> from the HBO show Silicon Valley. People would go to great lengths to cheat if the money/odds make it worthwhile.</p>",
      "rawMarkdown": "For a richer depiction, see a [clip][1] from the HBO show Silicon Valley. People would go to great lengths to cheat if the money/odds make it worthwhile.\n\n\n  [1]: https://www.youtube.com/watch?v=Y-W0CBOGnnI",
      "votes": null
    },
    {
      "id": "316123",
      "postDate": "04/18/2018 08:15:11",
      "content": "<p>Damn! Where is this world heading towards? Shameful indeed. \nBut the good thing is ML is helping fighting such cases.</p>",
      "rawMarkdown": "Damn! Where is this world heading towards? Shameful indeed. \nBut the good thing is ML is helping fighting such cases.",
      "votes": null
    },
    {
      "id": "316193",
      "postDate": "04/18/2018 11:46:48",
      "content": "<p>Every time you define incentives people will follow it, but sometimes in undesired ways.</p>",
      "rawMarkdown": "Every time you define incentives people will follow it, but sometimes in undesired ways.",
      "votes": null
    },
    {
      "id": "316232",
      "postDate": "04/18/2018 13:48:12",
      "content": "<p>Maybe they are giving the wrong incentive. Flip the discussion from detecting Fraud to measuring Click Quality. Change the payout to the channels from $/click to $/download = $/click x Click Quality.</p>\n\n<p>The channels probably serve all the ads. Give them the ability to measure Click Quality and to detect Fraud. Incentivise and equip the channels to police themselves.</p>\n\n<p>This aligns the interests of the parties involved (Channels and App Developers). Good business follows when interests are aligned. Click Quality is infinitely easier to measure than Fraud.</p>",
      "rawMarkdown": "Maybe they are giving the wrong incentive. Flip the discussion from detecting Fraud to measuring Click Quality. Change the payout to the channels from $/click to $/download = $/click x Click Quality.\n\nThe channels probably serve all the ads. Give them the ability to measure Click Quality and to detect Fraud. Incentivise and equip the channels to police themselves.\n\nThis aligns the interests of the parties involved (Channels and App Developers). Good business follows when interests are aligned. Click Quality is infinitely easier to measure than Fraud.",
      "votes": null
    },
    {
      "id": "316623",
      "postDate": "04/19/2018 14:19:47",
      "content": "<p>I believe this is the setup kaggle uses to run kernels ;)</p>",
      "rawMarkdown": "I believe this is the setup kaggle uses to run kernels ;)",
      "votes": null
    },
    {
      "id": "316695",
      "postDate": "04/19/2018 17:23:55",
      "content": "<p>So that's what's behind the <code>Anyone with an internet connection can make $$$ from home in your spare time</code> advertising ;)</p>",
      "rawMarkdown": "So that's what's behind the `Anyone with an internet connection can make $$$ from home in your spare time` advertising ;)",
      "votes": null
    },
    {
      "id": "316704",
      "postDate": "04/19/2018 17:46:16",
      "content": "<p>Is a CPA model (Cost-Per-Action) not legally/technically feasible in China ?</p>",
      "rawMarkdown": "Is a CPA model (Cost-Per-Action) not legally/technically feasible in China ?",
      "votes": null
    },
    {
      "id": "316778",
      "postDate": "04/19/2018 21:39:42",
      "content": "<p>Awesome, thanks for sharing these images. It helps me wrap my head around the actual business problem we are trying to solve, as opposed to it being just a bunch of abstraction. </p>",
      "rawMarkdown": "Awesome, thanks for sharing these images. It helps me wrap my head around the actual business problem we are trying to solve, as opposed to it being just a bunch of abstraction.",
      "votes": null
    },
    {
      "id": "316780",
      "postDate": "04/19/2018 21:56:15",
      "content": "<p>probably :) in fact there is a whole industry behind this with clear price for certain performance targets - I plan to follow up with another post explaining more details in the financial side of things :)</p>",
      "rawMarkdown": "probably :) in fact there is a whole industry behind this with clear price for certain performance targets - I plan to follow up with another post explaining more details in the financial side of things :)",
      "votes": null
    },
    {
      "id": "316783",
      "postDate": "04/19/2018 22:07:53",
      "content": "<p>There are phone emulators and programmable clickers as well. I wonder if those are used.</p>",
      "rawMarkdown": "There are phone emulators and programmable clickers as well. I wonder if those are used.",
      "votes": null
    },
    {
      "id": "317016",
      "postDate": "04/20/2018 13:25:06",
      "content": "<p>can you imagine doing this for 8 hours or more? I'd go dizzy after few hours constant clicking :)</p>",
      "rawMarkdown": "can you imagine doing this for 8 hours or more? I'd go dizzy after few hours constant clicking :)",
      "votes": null
    },
    {
      "id": "317390",
      "postDate": "04/21/2018 11:47:46",
      "content": "<p>Thank you very much !</p>",
      "rawMarkdown": "Thank you very much !",
      "votes": null
    },
    {
      "id": "319003",
      "postDate": "04/25/2018 02:41:33",
      "content": "<p>I know what you mean bro. In my country(China) the traffic policemen just wander along a street, write a note, paste to a car's window and the car owner will pay 200￥ or more.</p>",
      "rawMarkdown": "I know what you mean bro. In my country(China) the traffic policemen just wander along a street, write a note, paste to a car's window and the car owner will pay 200￥ or more.",
      "votes": null
    },
    {
      "id": "319004",
      "postDate": "04/25/2018 02:50:11",
      "content": "<p>I even doubt about why the competition host choose these basic features collected in training set. Maybe I think to analysis the add-link address and combine some simple features may perfectly solve this Fraud problem, the only factor made me involved in this competition is 184903891...</p>",
      "rawMarkdown": "I even doubt about why the competition host choose these basic features collected in training set. Maybe I think to analysis the add-link address and combine some simple features may perfectly solve this Fraud problem, the only factor made me involved in this competition is 184903891...",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 315797,
      "author_name": "stevenknguyen",
      "author_url": "",
      "post_date": "04/17/2018 16:01:04",
      "content": "<p>It's really crazy what people are willing to do. There was a story of a group of guys carrying around 400,000 unregistered sim cards a while back. It really sucks if you're trying to break out as app developer to compete against automated clicks. <a href=\"https://www.thestar.com.my/tech/tech-news/2017/06/13/thai-police-bust-400000-sim-card-chinese-click-farm/\">https://www.thestar.com.my/tech/tech-news/2017/06/13/thai-police-bust-400000-sim-card-chinese-click-farm/</a></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 315804,
      "author_name": "serigne",
      "author_url": "",
      "post_date": "04/17/2018 16:12:09",
      "content": "<p>Wow !..This is beyond crazy..:p</p>\n\n<p>Thanks Yifan for the sharing :)</p>",
      "votes": null,
      "replies": [
        {
          "id": 315829,
          "author_name": "yifanxie",
          "author_url": "",
          "post_date": "04/17/2018 16:34:57",
          "content": "<p>you are welcomed!  It really opened my eyes as well, even though I have heard about click farms before, seeing the images really hit home what kind of massive operation it is. </p>\n\n<p>I have more to share, but they are in Chinese so will take time to translate :)\nI guess I won't be seriously competing in this one, so getting a few discussion medals will do ^_^</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 315830,
      "author_name": "alexfir",
      "author_url": "",
      "post_date": "04/17/2018 16:35:49",
      "content": "<p>The article states that click farms can be used for app rating manipulation by downloading and installing apps. In this competition we use the fact of app downloads as a sign of real user. This means we have filtered/pre-processed data, otherwise we would see much more app downloads.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 316045,
      "author_name": "sijunhe9248",
      "author_url": "",
      "post_date": "04/18/2018 04:34:23",
      "content": "<p>For a richer depiction, see a <a href=\"https://www.youtube.com/watch?v=Y-W0CBOGnnI\">clip</a> from the HBO show Silicon Valley. People would go to great lengths to cheat if the money/odds make it worthwhile.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 316123,
      "author_name": "nuhsikander",
      "author_url": "",
      "post_date": "04/18/2018 08:15:11",
      "content": "<p>Damn! Where is this world heading towards? Shameful indeed. \nBut the good thing is ML is helping fighting such cases.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 316193,
      "author_name": "lgmoneda",
      "author_url": "",
      "post_date": "04/18/2018 11:46:48",
      "content": "<p>Every time you define incentives people will follow it, but sometimes in undesired ways.</p>",
      "votes": null,
      "replies": [
        {
          "id": 316232,
          "author_name": "johnoliverjones",
          "author_url": "",
          "post_date": "04/18/2018 13:48:12",
          "content": "<p>Maybe they are giving the wrong incentive. Flip the discussion from detecting Fraud to measuring Click Quality. Change the payout to the channels from $/click to $/download = $/click x Click Quality.</p>\n\n<p>The channels probably serve all the ads. Give them the ability to measure Click Quality and to detect Fraud. Incentivise and equip the channels to police themselves.</p>\n\n<p>This aligns the interests of the parties involved (Channels and App Developers). Good business follows when interests are aligned. Click Quality is infinitely easier to measure than Fraud.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 316623,
      "author_name": "scirpus",
      "author_url": "",
      "post_date": "04/19/2018 14:19:47",
      "content": "<p>I believe this is the setup kaggle uses to run kernels ;)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 316695,
      "author_name": "pliptor",
      "author_url": "",
      "post_date": "04/19/2018 17:23:55",
      "content": "<p>So that's what's behind the <code>Anyone with an internet connection can make $$$ from home in your spare time</code> advertising ;)</p>",
      "votes": null,
      "replies": [
        {
          "id": 316780,
          "author_name": "yifanxie",
          "author_url": "",
          "post_date": "04/19/2018 21:56:15",
          "content": "<p>probably :) in fact there is a whole industry behind this with clear price for certain performance targets - I plan to follow up with another post explaining more details in the financial side of things :)</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 316783,
          "author_name": "pliptor",
          "author_url": "",
          "post_date": "04/19/2018 22:07:53",
          "content": "<p>There are phone emulators and programmable clickers as well. I wonder if those are used.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 316704,
      "author_name": "ericperbos",
      "author_url": "",
      "post_date": "04/19/2018 17:46:16",
      "content": "<p>Is a CPA model (Cost-Per-Action) not legally/technically feasible in China ?</p>",
      "votes": null,
      "replies": [
        {
          "id": 319003,
          "author_name": "mcggood",
          "author_url": "",
          "post_date": "04/25/2018 02:41:33",
          "content": "<p>I know what you mean bro. In my country(China) the traffic policemen just wander along a street, write a note, paste to a car's window and the car owner will pay 200￥ or more.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 316778,
      "author_name": "attackgnome",
      "author_url": "",
      "post_date": "04/19/2018 21:39:42",
      "content": "<p>Awesome, thanks for sharing these images. It helps me wrap my head around the actual business problem we are trying to solve, as opposed to it being just a bunch of abstraction. </p>",
      "votes": null,
      "replies": [
        {
          "id": 319004,
          "author_name": "mcggood",
          "author_url": "",
          "post_date": "04/25/2018 02:50:11",
          "content": "<p>I even doubt about why the competition host choose these basic features collected in training set. Maybe I think to analysis the add-link address and combine some simple features may perfectly solve this Fraud problem, the only factor made me involved in this competition is 184903891...</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 317016,
      "author_name": "rajibbahar",
      "author_url": "",
      "post_date": "04/20/2018 13:25:06",
      "content": "<p>can you imagine doing this for 8 hours or more? I'd go dizzy after few hours constant clicking :)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 317390,
      "author_name": "pfichou",
      "author_url": "",
      "post_date": "04/21/2018 11:47:46",
      "content": "<p>Thank you very much !</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "315752": "Hello all,   I am supporting students from [DiveIntoCode][1] to learn practical (and impractical) ML skills via Kaggle.  so it is always nice to start tackling a machine learning problem by understanding its business context. Furthermore, seems no one has gone to explore the business context of click fraud here, and I have come across this nice article about chinese click farm.\n\nSo I hope by sharing this, you can get a better understanding of how these fake clicks are being generated, and who are doing it.  See below for a few interesting images:\n\n![enter image description here][2]\n\n![enter image description here][3]\n\nThe full link is here: https://kotaku.com/inside-chinese-click-farms-1795287821\n\n\n  [1]: https://diveintocode.jp/ai_curriculum\n  [2]: https://snag.gy/oiyHMS.jpg\n  [3]: https://snag.gy/TYfjq5.jpg",
    "315797": "It's really crazy what people are willing to do. There was a story of a group of guys carrying around 400,000 unregistered sim cards a while back. It really sucks if you're trying to break out as app developer to compete against automated clicks. https://www.thestar.com.my/tech/tech-news/2017/06/13/thai-police-bust-400000-sim-card-chinese-click-farm/",
    "315804": "Wow !..This is beyond crazy..:p\n\nThanks Yifan for the sharing :)",
    "315829": "you are welcomed!  It really opened my eyes as well, even though I have heard about click farms before, seeing the images really hit home what kind of massive operation it is. \n\nI have more to share, but they are in Chinese so will take time to translate :)\nI guess I won't be seriously competing in this one, so getting a few discussion medals will do ^_^",
    "315830": "The article states that click farms can be used for app rating manipulation by downloading and installing apps. In this competition we use the fact of app downloads as a sign of real user. This means we have filtered/pre-processed data, otherwise we would see much more app downloads.",
    "316045": "For a richer depiction, see a [clip][1] from the HBO show Silicon Valley. People would go to great lengths to cheat if the money/odds make it worthwhile.\n\n\n  [1]: https://www.youtube.com/watch?v=Y-W0CBOGnnI",
    "316123": "Damn! Where is this world heading towards? Shameful indeed. \nBut the good thing is ML is helping fighting such cases.",
    "316193": "Every time you define incentives people will follow it, but sometimes in undesired ways.",
    "316232": "Maybe they are giving the wrong incentive. Flip the discussion from detecting Fraud to measuring Click Quality. Change the payout to the channels from $/click to $/download = $/click x Click Quality.\n\nThe channels probably serve all the ads. Give them the ability to measure Click Quality and to detect Fraud. Incentivise and equip the channels to police themselves.\n\nThis aligns the interests of the parties involved (Channels and App Developers). Good business follows when interests are aligned. Click Quality is infinitely easier to measure than Fraud.",
    "316623": "I believe this is the setup kaggle uses to run kernels ;)",
    "316695": "So that's what's behind the `Anyone with an internet connection can make $$$ from home in your spare time` advertising ;)",
    "316704": "Is a CPA model (Cost-Per-Action) not legally/technically feasible in China ?",
    "316778": "Awesome, thanks for sharing these images. It helps me wrap my head around the actual business problem we are trying to solve, as opposed to it being just a bunch of abstraction.",
    "316780": "probably :) in fact there is a whole industry behind this with clear price for certain performance targets - I plan to follow up with another post explaining more details in the financial side of things :)",
    "316783": "There are phone emulators and programmable clickers as well. I wonder if those are used.",
    "317016": "can you imagine doing this for 8 hours or more? I'd go dizzy after few hours constant clicking :)",
    "317390": "Thank you very much !",
    "319003": "I know what you mean bro. In my country(China) the traffic policemen just wander along a street, write a note, paste to a car's window and the car owner will pay 200￥ or more.",
    "319004": "I even doubt about why the competition host choose these basic features collected in training set. Maybe I think to analysis the add-link address and combine some simple features may perfectly solve this Fraud problem, the only factor made me involved in this competition is 184903891..."
  },
  "source": "meta"
}