{
  "id": 321780,
  "title": "Recommending the most expensive items",
  "url": "/competitions/h-and-m-personalized-fashion-recommendations/discussion/321780",
  "author_name": "",
  "post_date": "2022-04-28T15:27:01.469264300Z",
  "votes": 1,
  "comment_count": 7,
  "views": 0,
  "content": "<p>Just for curiosity I recommended the most expensive items with a logic like this: </p>\n<p>if a customer X has purchased 20 items, most expensive 12 of her purchases will be recommended again.<br>\nif a customer Y has purchased 5 items, those 5 AND 7 of the overall most expensive 7 across all customers will be recommended.<br>\nif a customer Z has purchased 0 items, 12 of the overall most expensive 7 across all customers will be recommended.</p>\n<p>Got  a score of .0072. (Recommending most economical 12 was even worse)</p>\n<p>Any other ideas for using Price? I can think of price range of each customer, what else?</p>",
  "messages": [
    {
      "id": "1770785",
      "postDate": "04/28/2022 15:27:01",
      "content": "<p>Just for curiosity I recommended the most expensive items with a logic like this: </p>\n<p>if a customer X has purchased 20 items, most expensive 12 of her purchases will be recommended again.<br>\nif a customer Y has purchased 5 items, those 5 AND 7 of the overall most expensive 7 across all customers will be recommended.<br>\nif a customer Z has purchased 0 items, 12 of the overall most expensive 7 across all customers will be recommended.</p>\n<p>Got  a score of .0072. (Recommending most economical 12 was even worse)</p>\n<p>Any other ideas for using Price? I can think of price range of each customer, what else?</p>",
      "rawMarkdown": "Just for curiosity I recommended the most expensive items with a logic like this: \n\nif a customer X has purchased 20 items, most expensive 12 of her purchases will be recommended again.\nif a customer Y has purchased 5 items, those 5 AND 7 of the overall most expensive 7 across all customers will be recommended.\nif a customer Z has purchased 0 items, 12 of the overall most expensive 7 across all customers will be recommended.\n\nGot  a score of .0072. (Recommending most economical 12 was even worse)\n\nAny other ideas for using Price? I can think of price range of each customer, what else?",
      "votes": null
    },
    {
      "id": "1771081",
      "postDate": "04/28/2022 21:42:27",
      "content": "<p>interesting one</p>",
      "rawMarkdown": "interesting one",
      "votes": null
    },
    {
      "id": "1771112",
      "postDate": "04/28/2022 22:40:19",
      "content": "<p>Cheaper items usually sell more.</p>",
      "rawMarkdown": "Cheaper items usually sell more.",
      "votes": null
    },
    {
      "id": "1771121",
      "postDate": "04/28/2022 23:02:46",
      "content": "<p>I worked in e-commerce company in which I had access to ebay sales and I analyzed this. It is true that the probability of sales decreases with price but it is not linear. Two times more expensive items wasn't sold two times less frequently but something like 1.8 less frequently. So you actually earn more selling more expensive items.</p>",
      "rawMarkdown": "I worked in e-commerce company in which I had access to ebay sales and I analyzed this. It is true that the probability of sales decreases with price but it is not linear. Two times more expensive items wasn't sold two times less frequently but something like 1.8 less frequently. So you actually earn more selling more expensive items.",
      "votes": null
    },
    {
      "id": "1771251",
      "postDate": "04/29/2022 03:23:37",
      "content": "<p>I thought so and tried recommending the cheapest 12. That got an even worse score. </p>\n<p>I think most economical items are not most popular because they somehow implicitly signal they are bad in someway or the other and thus less people buy them.</p>\n<p>Anyways they also make less money for the seller. So the seller will also not be very enthusiastic about showing them to their customers.</p>\n<p>I will get more serious and look at other attributes as well😀</p>",
      "rawMarkdown": "I thought so and tried recommending the cheapest 12. That got an even worse score. \n\nI think most economical items are not most popular because they somehow implicitly signal they are bad in someway or the other and thus less people buy them.\n\nAnyways they also make less money for the seller. So the seller will also not be very enthusiastic about showing them to their customers.\n\nI will get more serious and look at other attributes as well😀",
      "votes": null
    },
    {
      "id": "1771636",
      "postDate": "04/29/2022 12:31:43",
      "content": "<p>According to my observation, demand and advertising often affect sales, not the cheaper the more. It may be a good strategy to recommend the range of higher-priced items that consumers buy, but the specific range needs further experimentation.</p>",
      "rawMarkdown": "According to my observation, demand and advertising often affect sales, not the cheaper the more. It may be a good strategy to recommend the range of higher-priced items that consumers buy, but the specific range needs further experimentation.",
      "votes": null
    },
    {
      "id": "1773095",
      "postDate": "04/30/2022 21:53:59",
      "content": "<p>This is a really neat approach! I would recommend maybe using the median price instead of the mean price, as that might be a more representative value for each customer. Additionally, you could try using a combination of the median and mean prices to get an even better recommendation.</p>",
      "rawMarkdown": "This is a really neat approach! I would recommend maybe using the median price instead of the mean price, as that might be a more representative value for each customer. Additionally, you could try using a combination of the median and mean prices to get an even better recommendation.",
      "votes": null
    },
    {
      "id": "1774002",
      "postDate": "05/01/2022 17:22:05",
      "content": "<p>When predicting for the LB, we don't know what the price of articles will be for that week.</p>\n<p>What we do know is the average price the articles sold for in the week before.<br>\nWe can also calculate for customers, for their previous purchases, what the articles' average price was for the weeks before they bought them.</p>\n<p>If you add both those features to your ranking model, it may help.<br>\n(I haven't tried yet, but I have found that there's correlation in these features for customer purchases)</p>",
      "rawMarkdown": "When predicting for the LB, we don't know what the price of articles will be for that week.\n\nWhat we do know is the average price the articles sold for in the week before.\nWe can also calculate for customers, for their previous purchases, what the articles' average price was for the weeks before they bought them.\n\nIf you add both those features to your ranking model, it may help.\n(I haven't tried yet, but I have found that there's correlation in these features for customer purchases)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1771081,
      "author_name": "vovastojoc",
      "author_url": "",
      "post_date": "04/28/2022 21:42:27",
      "content": "<p>interesting one</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1771112,
      "author_name": "jacob34",
      "author_url": "",
      "post_date": "04/28/2022 22:40:19",
      "content": "<p>Cheaper items usually sell more.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1771121,
          "author_name": "paweljankiewicz",
          "author_url": "",
          "post_date": "04/28/2022 23:02:46",
          "content": "<p>I worked in e-commerce company in which I had access to ebay sales and I analyzed this. It is true that the probability of sales decreases with price but it is not linear. Two times more expensive items wasn't sold two times less frequently but something like 1.8 less frequently. So you actually earn more selling more expensive items.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1771251,
          "author_name": "srinivassateesh",
          "author_url": "",
          "post_date": "04/29/2022 03:23:37",
          "content": "<p>I thought so and tried recommending the cheapest 12. That got an even worse score. </p>\n<p>I think most economical items are not most popular because they somehow implicitly signal they are bad in someway or the other and thus less people buy them.</p>\n<p>Anyways they also make less money for the seller. So the seller will also not be very enthusiastic about showing them to their customers.</p>\n<p>I will get more serious and look at other attributes as well😀</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1771636,
      "author_name": "",
      "author_url": "",
      "post_date": "04/29/2022 12:31:43",
      "content": "<p>According to my observation, demand and advertising often affect sales, not the cheaper the more. It may be a good strategy to recommend the range of higher-priced items that consumers buy, but the specific range needs further experimentation.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1773095,
      "author_name": "",
      "author_url": "",
      "post_date": "04/30/2022 21:53:59",
      "content": "<p>This is a really neat approach! I would recommend maybe using the median price instead of the mean price, as that might be a more representative value for each customer. Additionally, you could try using a combination of the median and mean prices to get an even better recommendation.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1774002,
      "author_name": "jacob34",
      "author_url": "",
      "post_date": "05/01/2022 17:22:05",
      "content": "<p>When predicting for the LB, we don't know what the price of articles will be for that week.</p>\n<p>What we do know is the average price the articles sold for in the week before.<br>\nWe can also calculate for customers, for their previous purchases, what the articles' average price was for the weeks before they bought them.</p>\n<p>If you add both those features to your ranking model, it may help.<br>\n(I haven't tried yet, but I have found that there's correlation in these features for customer purchases)</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1770785": "Just for curiosity I recommended the most expensive items with a logic like this: \n\nif a customer X has purchased 20 items, most expensive 12 of her purchases will be recommended again.\nif a customer Y has purchased 5 items, those 5 AND 7 of the overall most expensive 7 across all customers will be recommended.\nif a customer Z has purchased 0 items, 12 of the overall most expensive 7 across all customers will be recommended.\n\nGot  a score of .0072. (Recommending most economical 12 was even worse)\n\nAny other ideas for using Price? I can think of price range of each customer, what else?",
    "1771081": "interesting one",
    "1771112": "Cheaper items usually sell more.",
    "1771121": "I worked in e-commerce company in which I had access to ebay sales and I analyzed this. It is true that the probability of sales decreases with price but it is not linear. Two times more expensive items wasn't sold two times less frequently but something like 1.8 less frequently. So you actually earn more selling more expensive items.",
    "1771251": "I thought so and tried recommending the cheapest 12. That got an even worse score. \n\nI think most economical items are not most popular because they somehow implicitly signal they are bad in someway or the other and thus less people buy them.\n\nAnyways they also make less money for the seller. So the seller will also not be very enthusiastic about showing them to their customers.\n\nI will get more serious and look at other attributes as well😀",
    "1771636": "According to my observation, demand and advertising often affect sales, not the cheaper the more. It may be a good strategy to recommend the range of higher-priced items that consumers buy, but the specific range needs further experimentation.",
    "1773095": "This is a really neat approach! I would recommend maybe using the median price instead of the mean price, as that might be a more representative value for each customer. Additionally, you could try using a combination of the median and mean prices to get an even better recommendation.",
    "1774002": "When predicting for the LB, we don't know what the price of articles will be for that week.\n\nWhat we do know is the average price the articles sold for in the week before.\nWe can also calculate for customers, for their previous purchases, what the articles' average price was for the weeks before they bought them.\n\nIf you add both those features to your ranking model, it may help.\n(I haven't tried yet, but I have found that there's correlation in these features for customer purchases)"
  },
  "source": "meta"
}