{
  "id": 324112,
  "title": "Some insights from scraping the H&M online shop",
  "url": "/competitions/h-and-m-personalized-fashion-recommendations/discussion/324112",
  "author_name": "",
  "post_date": "2022-05-10T07:16:18.193140800Z",
  "votes": 3,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi all,</p>\n<p>congrats to the winning solutions - great job everyone.</p>\n<p>As some wondered if the H&amp;M online shop would be any useful, I wanted to share some information regarding the information:</p>\n<p>I scraped the following information from the web shops for different countries:</p>\n<ul>\n<li>number of ratings of a product</li>\n<li>average rating of a product</li>\n<li>number of reviews of a product</li>\n<li>categorization of articles (age_gender, e.g. teen_male)</li>\n<li>Recommendations (others bought also, style with)</li>\n</ul>\n<p>Using the data and some facebook/instagram research, I think that the data for the competition is from the Netherlands as it nicely fits the following:</p>\n<ul>\n<li>prices seem to be in Euro</li>\n<li>sales with 20% off on April 11th 2020: <a href=\"https://www.facebook.com/hmnetherlands/photos/a.785601701506054/2881542961911907\" target=\"_blank\">https://www.facebook.com/hmnetherlands/photos/a.785601701506054/2881542961911907</a></li>\n<li>stores closed starting at March 19th 2020, 17h: <a href=\"https://www.facebook.com/hmnetherlands/photos/a.785601701506054/2834092859990251/\" target=\"_blank\">https://www.facebook.com/hmnetherlands/photos/a.785601701506054/2834092859990251/</a></li>\n<li>highest correlation of ratings with actual sales</li>\n</ul>\n<p>What's interesting: there was a 20% sale in the prediction week:<br>\n<a href=\"https://www.facebook.com/hmnetherlands/photos/a.785601701506054/3346515532081312\" target=\"_blank\">https://www.facebook.com/hmnetherlands/photos/a.785601701506054/3346515532081312</a></p>\n<p>I had technical problems towards the end and could not submit a good prediction, but still some preliminary findings:</p>\n<ul>\n<li>the recommendations were not useful and could not increase coverage of the purchased articles compared to best selling articles, articles previous purchased, etc.</li>\n<li>prices could be helpful for the cold start articles, but these are hard to incorporate anyway</li>\n<li>average ratings and number of reviews did not show much correlation with the sales data</li>\n<li>number of ratings seemed to be really good (top 6 feature in my latest model)</li>\n<li>the suspicion about the sale in the prediction week still lingers on my todo list</li>\n</ul>",
  "messages": [
    {
      "id": "1783179",
      "postDate": "05/10/2022 07:16:18",
      "content": "<p>Hi all,</p>\n<p>congrats to the winning solutions - great job everyone.</p>\n<p>As some wondered if the H&amp;M online shop would be any useful, I wanted to share some information regarding the information:</p>\n<p>I scraped the following information from the web shops for different countries:</p>\n<ul>\n<li>number of ratings of a product</li>\n<li>average rating of a product</li>\n<li>number of reviews of a product</li>\n<li>categorization of articles (age_gender, e.g. teen_male)</li>\n<li>Recommendations (others bought also, style with)</li>\n</ul>\n<p>Using the data and some facebook/instagram research, I think that the data for the competition is from the Netherlands as it nicely fits the following:</p>\n<ul>\n<li>prices seem to be in Euro</li>\n<li>sales with 20% off on April 11th 2020: <a href=\"https://www.facebook.com/hmnetherlands/photos/a.785601701506054/2881542961911907\" target=\"_blank\">https://www.facebook.com/hmnetherlands/photos/a.785601701506054/2881542961911907</a></li>\n<li>stores closed starting at March 19th 2020, 17h: <a href=\"https://www.facebook.com/hmnetherlands/photos/a.785601701506054/2834092859990251/\" target=\"_blank\">https://www.facebook.com/hmnetherlands/photos/a.785601701506054/2834092859990251/</a></li>\n<li>highest correlation of ratings with actual sales</li>\n</ul>\n<p>What's interesting: there was a 20% sale in the prediction week:<br>\n<a href=\"https://www.facebook.com/hmnetherlands/photos/a.785601701506054/3346515532081312\" target=\"_blank\">https://www.facebook.com/hmnetherlands/photos/a.785601701506054/3346515532081312</a></p>\n<p>I had technical problems towards the end and could not submit a good prediction, but still some preliminary findings:</p>\n<ul>\n<li>the recommendations were not useful and could not increase coverage of the purchased articles compared to best selling articles, articles previous purchased, etc.</li>\n<li>prices could be helpful for the cold start articles, but these are hard to incorporate anyway</li>\n<li>average ratings and number of reviews did not show much correlation with the sales data</li>\n<li>number of ratings seemed to be really good (top 6 feature in my latest model)</li>\n<li>the suspicion about the sale in the prediction week still lingers on my todo list</li>\n</ul>",
      "rawMarkdown": "Hi all,\n\ncongrats to the winning solutions - great job everyone.\n\nAs some wondered if the H&M online shop would be any useful, I wanted to share some information regarding the information:\n\nI scraped the following information from the web shops for different countries:\n\n- number of ratings of a product\n- average rating of a product\n- number of reviews of a product\n- categorization of articles (age_gender, e.g. teen_male)\n- Recommendations (others bought also, style with)\n\nUsing the data and some facebook/instagram research, I think that the data for the competition is from the Netherlands as it nicely fits the following:\n\n- prices seem to be in Euro\n- sales with 20% off on April 11th 2020: https://www.facebook.com/hmnetherlands/photos/a.785601701506054/2881542961911907\n- stores closed starting at March 19th 2020, 17h: https://www.facebook.com/hmnetherlands/photos/a.785601701506054/2834092859990251/\n- highest correlation of ratings with actual sales\n\nWhat's interesting: there was a 20% sale in the prediction week:\nhttps://www.facebook.com/hmnetherlands/photos/a.785601701506054/3346515532081312\n\nI had technical problems towards the end and could not submit a good prediction, but still some preliminary findings:\n\n- the recommendations were not useful and could not increase coverage of the purchased articles compared to best selling articles, articles previous purchased, etc.\n- prices could be helpful for the cold start articles, but these are hard to incorporate anyway\n- average ratings and number of reviews did not show much correlation with the sales data\n- number of ratings seemed to be really good (top 6 feature in my latest model)\n- the suspicion about the sale in the prediction week still lingers on my todo list",
      "votes": null
    },
    {
      "id": "1783886",
      "postDate": "05/10/2022 18:53:10",
      "content": "<p>Thanks for posting this! It's really helpful to see what kind of data was available and how it might be used. I'm also curious about the sale in the prediction week - it would be interesting to know if that had any effect on the results.</p>",
      "rawMarkdown": "Thanks for posting this! It's really helpful to see what kind of data was available and how it might be used. I'm also curious about the sale in the prediction week - it would be interesting to know if that had any effect on the results.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1783886,
      "author_name": "",
      "author_url": "",
      "post_date": "05/10/2022 18:53:10",
      "content": "<p>Thanks for posting this! It's really helpful to see what kind of data was available and how it might be used. I'm also curious about the sale in the prediction week - it would be interesting to know if that had any effect on the results.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1783179": "Hi all,\n\ncongrats to the winning solutions - great job everyone.\n\nAs some wondered if the H&M online shop would be any useful, I wanted to share some information regarding the information:\n\nI scraped the following information from the web shops for different countries:\n\n- number of ratings of a product\n- average rating of a product\n- number of reviews of a product\n- categorization of articles (age_gender, e.g. teen_male)\n- Recommendations (others bought also, style with)\n\nUsing the data and some facebook/instagram research, I think that the data for the competition is from the Netherlands as it nicely fits the following:\n\n- prices seem to be in Euro\n- sales with 20% off on April 11th 2020: https://www.facebook.com/hmnetherlands/photos/a.785601701506054/2881542961911907\n- stores closed starting at March 19th 2020, 17h: https://www.facebook.com/hmnetherlands/photos/a.785601701506054/2834092859990251/\n- highest correlation of ratings with actual sales\n\nWhat's interesting: there was a 20% sale in the prediction week:\nhttps://www.facebook.com/hmnetherlands/photos/a.785601701506054/3346515532081312\n\nI had technical problems towards the end and could not submit a good prediction, but still some preliminary findings:\n\n- the recommendations were not useful and could not increase coverage of the purchased articles compared to best selling articles, articles previous purchased, etc.\n- prices could be helpful for the cold start articles, but these are hard to incorporate anyway\n- average ratings and number of reviews did not show much correlation with the sales data\n- number of ratings seemed to be really good (top 6 feature in my latest model)\n- the suspicion about the sale in the prediction week still lingers on my todo list",
    "1783886": "Thanks for posting this! It's really helpful to see what kind of data was available and how it might be used. I'm also curious about the sale in the prediction week - it would be interesting to know if that had any effect on the results."
  },
  "source": "meta"
}