{
  "id": 66862,
  "title": "How do you deal with gclID?",
  "url": "/competitions/ga-customer-revenue-prediction/discussion/66862",
  "author_name": "",
  "post_date": "2018-09-26T07:13:20.915424Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I wonder how do people extract useful info from gclID (<code>trafficSource.adwordsClickInfo.gclId</code>). </p>\n\n<p>Goolging around I found several posts describing what it is, for example, <a href=\"https://www.ppchero.com/what-the-gclid/\">this one</a>. And it seems to contain hashed/encrypted info that is available in other features. So do you do anything special about it? Any n-gram parsing to extract some patterns?</p>",
  "messages": [
    {
      "id": "393994",
      "postDate": "09/26/2018 07:13:20",
      "content": "<p>I wonder how do people extract useful info from gclID (<code>trafficSource.adwordsClickInfo.gclId</code>). </p>\n\n<p>Goolging around I found several posts describing what it is, for example, <a href=\"https://www.ppchero.com/what-the-gclid/\">this one</a>. And it seems to contain hashed/encrypted info that is available in other features. So do you do anything special about it? Any n-gram parsing to extract some patterns?</p>",
      "rawMarkdown": "I wonder how do people extract useful info from gclID (`trafficSource.adwordsClickInfo.gclId`). \n\nGoolging around I found several posts describing what it is, for example, [this one](https://www.ppchero.com/what-the-gclid/). And it seems to contain hashed/encrypted info that is available in other features. So do you do anything special about it? Any n-gram parsing to extract some patterns?",
      "votes": null
    },
    {
      "id": "400043",
      "postDate": "10/07/2018 12:52:38",
      "content": "<p>Unfortunately number of rows with non-null gcllD is too small... otherwise we could try extracting yet another timestamp from it.</p>",
      "rawMarkdown": "Unfortunately number of rows with non-null gcllD is too small... otherwise we could try extracting yet another timestamp from it.",
      "votes": null
    },
    {
      "id": "400191",
      "postDate": "10/07/2018 19:33:30",
      "content": "<p>The descriptions I find on glcID are quite ambiguous. Don't know where that post got the idea of encoding protocol buffers into the string, I suppose it could be possible. I would think it is more likely a string identifier linking this or other info. But maybe they found a better source of information. </p>\n\n<p>Some of the posts on the topic seem to refer to glcID being used as a way to monitor/track ad clicks and how they lead to conversions. So I tried to count how often each ID repeats (if it is used as unique ID to track sessions in relation to specific ad click etc). Then use the cumulative count over time for a feature. </p>\n\n<p>Doubt if it makes much of a difference, and it is hard to really understand what it is about. Much as most of the parameter details in this data, limited descriptions.</p>",
      "rawMarkdown": "The descriptions I find on glcID are quite ambiguous. Don't know where that post got the idea of encoding protocol buffers into the string, I suppose it could be possible. I would think it is more likely a string identifier linking this or other info. But maybe they found a better source of information. \n\nSome of the posts on the topic seem to refer to glcID being used as a way to monitor/track ad clicks and how they lead to conversions. So I tried to count how often each ID repeats (if it is used as unique ID to track sessions in relation to specific ad click etc). Then use the cumulative count over time for a feature. \n\nDoubt if it makes much of a difference, and it is hard to really understand what it is about. Much as most of the parameter details in this data, limited descriptions.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 400043,
      "author_name": "fingoldo",
      "author_url": "",
      "post_date": "10/07/2018 12:52:38",
      "content": "<p>Unfortunately number of rows with non-null gcllD is too small... otherwise we could try extracting yet another timestamp from it.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 400191,
      "author_name": "donkeys",
      "author_url": "",
      "post_date": "10/07/2018 19:33:30",
      "content": "<p>The descriptions I find on glcID are quite ambiguous. Don't know where that post got the idea of encoding protocol buffers into the string, I suppose it could be possible. I would think it is more likely a string identifier linking this or other info. But maybe they found a better source of information. </p>\n\n<p>Some of the posts on the topic seem to refer to glcID being used as a way to monitor/track ad clicks and how they lead to conversions. So I tried to count how often each ID repeats (if it is used as unique ID to track sessions in relation to specific ad click etc). Then use the cumulative count over time for a feature. </p>\n\n<p>Doubt if it makes much of a difference, and it is hard to really understand what it is about. Much as most of the parameter details in this data, limited descriptions.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "393994": "I wonder how do people extract useful info from gclID (`trafficSource.adwordsClickInfo.gclId`). \n\nGoolging around I found several posts describing what it is, for example, [this one](https://www.ppchero.com/what-the-gclid/). And it seems to contain hashed/encrypted info that is available in other features. So do you do anything special about it? Any n-gram parsing to extract some patterns?",
    "400043": "Unfortunately number of rows with non-null gcllD is too small... otherwise we could try extracting yet another timestamp from it.",
    "400191": "The descriptions I find on glcID are quite ambiguous. Don't know where that post got the idea of encoding protocol buffers into the string, I suppose it could be possible. I would think it is more likely a string identifier linking this or other info. But maybe they found a better source of information. \n\nSome of the posts on the topic seem to refer to glcID being used as a way to monitor/track ad clicks and how they lead to conversions. So I tried to count how often each ID repeats (if it is used as unique ID to track sessions in relation to specific ad click etc). Then use the cumulative count over time for a feature. \n\nDoubt if it makes much of a difference, and it is hard to really understand what it is about. Much as most of the parameter details in this data, limited descriptions."
  },
  "source": "meta"
}