Measure Most-Viewed URLs and Subscription Conversion Rates
Company: Vanta
Role: Software Engineer
Category: System Design
Difficulty: medium
Interview Round: Onsite
Design a metrics system for a web application that reports its most-viewed URLs and subscription conversion rate. Explain event definitions, data collection, aggregation, and how users can interpret the results correctly.
### Part 1 — Most-Viewed URLs
Define a view and URL identity, then describe how to compute top URLs over a requested time window.
#### What This Part Should Cover
URL normalization, raw versus unique views, duplicate events, time windows, and scalable count aggregation.
### Part 2 — Subscription Conversion
Define the conversion numerator, eligible denominator, attribution window, and user or session identity. Explain how the system joins events and handles delayed conversions.
#### What This Part Should Cover
A coherent funnel or cohort definition, deduplication, time alignment, and avoiding misleading rates.
### Constraints
Traffic scale, freshness, and exact business definitions are unspecified. Do not assume that every page view belongs in the conversion denominator or that a subscription event always belongs to the same day's traffic.
### Clarifying Questions
- Are URL views page loads, route changes, sessions, or distinct visitors?
- Which query parameters distinguish content, and which should be normalized away?
- Is conversion measured per visitor, account, session, or eligible cohort?
- How are bots, retries, cancellations, and late-arriving events treated?
```hint Fix the denominator before building the dashboard
A precise numerator cannot rescue a conversion rate whose eligible population is undefined.
```
### What a Strong Answer Covers
- Explicit metric definitions and a durable event model.
- Windowed top-URL aggregation and identity-aware conversion computation.
- Freshness, late data, correctness checks, and transparent reporting.
### Follow-up Questions
- How would a subscription several days after the first visit affect cohort reporting?
- How would you detect that a client release stopped sending view events?
Overview: Design web metrics around precise view, URL, identity, cohort, and conversion definitions with durable aggregation and late-event handling.
Measure Most-Viewed URLs and Subscription Conversion Rates
Vanta
Sep 12, 2026
mediumSoftware EngineerOnsiteSystem Design
0
0
Design a metrics system for a web application that reports its most-viewed URLs and subscription conversion rate. Explain event definitions, data collection, aggregation, and how users can interpret the results correctly.
Part 1 — Most-Viewed URLs
Define a view and URL identity, then describe how to compute top URLs over a requested time window.
What This Part Should Cover Guidance
URL normalization, raw versus unique views, duplicate events, time windows, and scalable count aggregation.
Part 2 — Subscription Conversion
Define the conversion numerator, eligible denominator, attribution window, and user or session identity. Explain how the system joins events and handles delayed conversions.
What This Part Should Cover Guidance
A coherent funnel or cohort definition, deduplication, time alignment, and avoiding misleading rates.
Constraints
Traffic scale, freshness, and exact business definitions are unspecified. Do not assume that every page view belongs in the conversion denominator or that a subscription event always belongs to the same day's traffic.
Clarifying Questions Guidance
Are URL views page loads, route changes, sessions, or distinct visitors?
Which query parameters distinguish content, and which should be normalized away?
Is conversion measured per visitor, account, session, or eligible cohort?
How are bots, retries, cancellations, and late-arriving events treated?
What a Strong Answer Covers Guidance
Explicit metric definitions and a durable event model.
Windowed top-URL aggregation and identity-aware conversion computation.
Freshness, late data, correctness checks, and transparent reporting.
Follow-up Questions Guidance
How would a subscription several days after the first visit affect cohort reporting?
How would you detect that a client release stopped sending view events?