Amazon · Statistics & Data Analysis

How would you test a price increase?

TrueInterview

October 7, 2026 · 2 min read

Question

Imagine you are a data scientist working on a B2C subscription product: an AI video editing tool that offers a free trial and several paid plans. Product leadership is weighing a price increase—for instance, raising the Pro plan's monthly price by 10-25%—and may also adjust what each plan includes. Outline the analysis and experimentation plan that would support this decision. Specifically:

Should we raise prices at all? Before running any test, how would you assess whether an increase is likely to help?
What exactly would you launch, and for whom? Consider the price change and/or packaging change—list price, included credits, tier structure, annual versus monthly billing—and specify which plans, regions, and customer segments it would affect.
How would you estimate price sensitivity / elasticity using existing data, and how much confidence would you place in that estimate?
Design the experiment. Cover the randomization unit, treatment arms, population, duration, and statistical power, and specifically explain how you would treat existing subscribers differently from new signups.
What are the success metrics? Include primary, diagnostic, and guardrail metrics, along with the tradeoffs among them.
Which segments matter, and why is segmentation important in this case? Examples include new versus existing users, region, creator/professional versus casual users, and plan tier.
What confounders, biases, and risks would you monitor? Examples include seasonality, competitor promotions, selection effects, plan switching and cannibalization, and delayed churn.
What would you do if full randomization of price is not possible due to legal, app store, or fairness constraints?
How would you interpret the results and make the decision? Include stopping criteria, a ramp plan, and the final rollout recommendation. Overview: This is an Amazon data scientist technical screen case on pricing. It asks you to decide whether to raise the price of a subscription product, select the price point and packaging, and design the experiment that validates the choice. A strong answer covers per-visitor profit and LTV metrics, guardrails, elasticity estimation, randomization at the billing-account level, and how to treat existing subscribers when they renew. It also addresses two common pitfalls: evaluating a subscription price test on short-horizon revenue, and measuring churn only among users who converted.

Amazon · Statistics & Data Analysis

How would you test a price increase?

TrueInterview

October 7, 2026 · 2 min read

Question

Should we raise prices at all? Before running any test, how would you assess whether an increase is likely to help?
What exactly would you launch, and for whom? Consider the price change and/or packaging change—list price, included credits, tier structure, annual versus monthly billing—and specify which plans, regions, and customer segments it would affect.
How would you estimate price sensitivity / elasticity using existing data, and how much confidence would you place in that estimate?
Design the experiment. Cover the randomization unit, treatment arms, population, duration, and statistical power, and specifically explain how you would treat existing subscribers differently from new signups.
What are the success metrics? Include primary, diagnostic, and guardrail metrics, along with the tradeoffs among them.
Which segments matter, and why is segmentation important in this case? Examples include new versus existing users, region, creator/professional versus casual users, and plan tier.
What confounders, biases, and risks would you monitor? Examples include seasonality, competitor promotions, selection effects, plan switching and cannibalization, and delayed churn.
What would you do if full randomization of price is not possible due to legal, app store, or fairness constraints?
How would you interpret the results and make the decision? Include stopping criteria, a ramp plan, and the final rollout recommendation. Overview: This is an Amazon data scientist technical screen case on pricing. It asks you to decide whether to raise the price of a subscription product, select the price point and packaging, and design the experiment that validates the choice. A strong answer covers per-visitor profit and LTV metrics, guardrails, elasticity estimation, randomization at the billing-account level, and how to treat existing subscribers when they renew. It also addresses two common pitfalls: evaluating a subscription price test on short-horizon revenue, and measuring churn only among users who converted.