引導式測驗 · 中等 等級
人工智能预测和减少回报测验
測試人工智慧退貨管理:結果定義、退貨原因、產品修復、隱私、校準、幹預測試、客戶影響和有效退貨
問題 1 的 8
A product receives a high return-risk score. What does the score establish?
預覽每個問題8 問題
當您參加測驗時,就會出現答案和解釋。
- A product receives a high return-risk score. What does the score establish?
- That the customer plans to abuse return policies
- That the product is defective
- That the model predicts elevated return likelihood under its learned data, not why a specific customer will return
- That the retailer should block the order
- Why separate fit-related returns from late delivery and damage?
- Different causes require different fixes and can be hidden inside one return label
- All return reasons respond to size charts
- A single combined label improves diagnosis
- Return cause never matters to analysis
- Which action addresses a recurring fit problem without penalizing shoppers?
- Restrict purchases by customers with past returns
- Hide the item from search
- Charge higher return fees only to one risk group
- Improve measurements, size charts, and fit descriptions for the affected item
- What does high predictive importance of a customer feature establish?
- It proves the feature causes returns
- It helps the model predict in its data but does not by itself justify using it for an intervention
- It guarantees a fair business decision
- It means the feature is legally approved
- How should a team test a packaging change intended to reduce damage returns?
- Deploy everywhere and attribute any change to packaging
- Measure only clicks on the product page
- Compare the intervention with a credible baseline and measure return reasons and customer outcomes
- Use the training score as the outcome
- Why can restricting high-risk orders create a feedback loop?
- The score remains independent of all business actions
- Returns cannot be measured after a restriction
- It automatically improves calibration
- The policy changes later data, so the model learns from its own intervention
- Which test split better measures performance on future orders?
- A later time period kept out of model training
- A random split with same orders in train and test
- Only training data
- The highest-risk orders only
- Which outcome should be tracked alongside return rate?
- Only the number of risk scores generated
- Customer satisfaction, complaints, conversion, and valid-return completion
- The model’s vocabulary size
- The number of warehouse employees