# Investigate a Decrease in Webpage Warnings
A system that warns users about malicious webpages shows a decrease in the number of warnings displayed. How would you investigate? Explain how you would distinguish an improvement in safety from a detection, delivery, or measurement problem, and what evidence would guide the next action.
### What a Strong Answer Covers
- Validation of the warning metric, time comparison, and logging completeness.
- A decomposition from browsing traffic through scoring and warning display.
- Segments and release checks that distinguish prevalence changes from pipeline faults.
- Independent evidence of malicious exposure and a decision appropriate to the cause.
```hint Follow a browsing encounter
A malicious page can be missed because it was never scored, not flagged, not rendered, or not logged.
```
### Follow-up Questions
- What would you conclude if flags stayed constant but displayed warnings fell?
- How would you estimate whether false negatives increased?
Overview: Diagnose falling malicious-webpage warnings by decomposing traffic, scoring, flags, rendering, and labels to separate improvement from failure.
A system that warns users about malicious webpages shows a decrease in the number of warnings displayed. How would you investigate? Explain how you would distinguish an improvement in safety from a detection, delivery, or measurement problem, and what evidence would guide the next action.
What a Strong Answer Covers Guidance
Validation of the warning metric, time comparison, and logging completeness.
A decomposition from browsing traffic through scoring and warning display.
Segments and release checks that distinguish prevalence changes from pipeline faults.
Independent evidence of malicious exposure and a decision appropriate to the cause.
Follow-up Questions Guidance
What would you conclude if flags stayed constant but displayed warnings fell?
How would you estimate whether false negatives increased?