Refer to the exhibit: Which statement is true, based on the plots above?

Question

Options

• AApproximately twice as many customers with the top ten percent of predicted probabilities are expected to have a positive predictive event.
• BApproximately ten percent of a randomly selected subset of twenty percent of the customers are expected to have a positive predictive event.
• CApproximately twenty percent of the customers with a predicted score of 3 have a positive predicted class.
• DApproximately ten percent of those customers with the top twenty percent of predicted probabilities are expected to have a positive predictive event.

• A
  49% (26)
• B
  28% (15)
• C
  8% (4)
• D
  15% (8)

Explanation

Option A correctly describes a lift of 2.0 at the top decile, which is a standard reading from a lift chart: when you rank customers by predicted probability and target the top 10%, you capture approximately twice as many positive events as you would by randomly selecting 10% of the population. This demonstrates the model's ability to concentrate positive cases at the high end of the score distribution.

Why the distractors are wrong:

• B confuses the baseline random response rate with a conditional probability - randomly selecting 20% of customers yields roughly 20% of all positive events, not 10% of that subset.
• C misreads the chart type; lift/gains charts plot cumulative population percentiles, not individual score values like "3," so no such direct reading exists.
• D inverts the logic - if only 10% of the top 20% had positive events, the model would perform at or below random chance, defeating the purpose of the model entirely.

Memory tip: Think of lift as a multiplier. A lift of 2 at the top 10% means "2× better than random." If you see a question asking about top-percentage groups and comparing to random selection, you're reading a lift chart - the answer will almost always highlight that the model outperforms random targeting.

No community discussion yet for this question.