{
  "id": 746457,
  "title": "Clarification on the target window and prediction cutoff date",
  "url": "/competitions/global-e-commerce-customer-purchase-prediction/discussion/746457",
  "author_name": "CAN WEI28",
  "post_date": "2026-10-07T06:51:23.650000",
  "votes": 0,
  "comment_count": 0,
  "views": null,
  "content": "<p>Hi,</p>\n<p>Thank you for sharing this dataset. We are using it for a university group project and would like to clarify the time definition of the target.</p>\n<ol>\n<li><p>What is the prediction cutoff date for the target in train.csv? Does target = 1 mean that the customer made at least one purchase during the 30 days immediately after this cutoff?</p></li>\n<li><p>Are all records in transactions.csv and sessions.csv from before that cutoff? Were customer attributes such as loyalty_score and customer_tier also measured before the cutoff?</p></li>\n<li><p>We are considering using the day after the latest observed transaction/session date as the reference date for Recency features. Is this consistent with the intended prediction setup, or should we use a different date?</p></li>\n</ol>\n<p>Do train and test customers share the same cutoff date and 30-day target window?</p>\n<p>We want to make sure our features use only information available before the prediction window.</p>\n<p>Thank you for your help!</p>",
  "messages": [
    {
      "id": 3534016,
      "postDate": "2026-10-07T06:51:23.650Z",
      "content": "<p>Hi,</p>\n<p>Thank you for sharing this dataset. We are using it for a university group project and would like to clarify the time definition of the target.</p>\n<ol>\n<li><p>What is the prediction cutoff date for the target in train.csv? Does target = 1 mean that the customer made at least one purchase during the 30 days immediately after this cutoff?</p></li>\n<li><p>Are all records in transactions.csv and sessions.csv from before that cutoff? Were customer attributes such as loyalty_score and customer_tier also measured before the cutoff?</p></li>\n<li><p>We are considering using the day after the latest observed transaction/session date as the reference date for Recency features. Is this consistent with the intended prediction setup, or should we use a different date?</p></li>\n</ol>\n<p>Do train and test customers share the same cutoff date and 30-day target window?</p>\n<p>We want to make sure our features use only information available before the prediction window.</p>\n<p>Thank you for your help!</p>",
      "rawMarkdown": "Hi,\n\nThank you for sharing this dataset. We are using it for a university group project and would like to clarify the time definition of the target.\n\n1. What is the prediction cutoff date for the target in train.csv? Does target = 1 mean that the customer made at least one purchase during the 30 days immediately after this cutoff?\n\n2. Are all records in transactions.csv and sessions.csv from before that cutoff? Were customer attributes such as loyalty_score and customer_tier also measured before the cutoff?\n\n3. We are considering using the day after the latest observed transaction/session date as the reference date for Recency features. Is this consistent with the intended prediction setup, or should we use a different date?\n\nDo train and test customers share the same cutoff date and 30-day target window?\n\nWe want to make sure our features use only information available before the prediction window.\n\nThank you for your help!"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "3534016": "Hi,\n\nThank you for sharing this dataset. We are using it for a university group project and would like to clarify the time definition of the target.\n\n1. What is the prediction cutoff date for the target in train.csv? Does target = 1 mean that the customer made at least one purchase during the 30 days immediately after this cutoff?\n\n2. Are all records in transactions.csv and sessions.csv from before that cutoff? Were customer attributes such as loyalty_score and customer_tier also measured before the cutoff?\n\n3. We are considering using the day after the latest observed transaction/session date as the reference date for Recency features. Is this consistent with the intended prediction setup, or should we use a different date?\n\nDo train and test customers share the same cutoff date and 30-day target window?\n\nWe want to make sure our features use only information available before the prediction window.\n\nThank you for your help!"
  }
}