Design an experiment to evaluate an onboarding progress bar

Scenario

In a QuickBooks-like onboarding flow, the product team adds a progress bar.

The onboarding has 4 steps:

1. Onboard
2. Send payment
3. Set up bank account
4. File the tax

Question

How would you evaluate whether the progress bar helps?

Your answer should include:

• Hypothesis and what “helps” means
• Primary metric + diagnostic metrics + guardrails (with tradeoffs)
• Experiment design (randomization unit, sample size/power considerations at a high level, duration)
• Key pitfalls (novelty effects, instrumentation, segment heterogeneity) and how you’d address them