# Product Execution Interview Example: Increase Disney+ Retention

Aced (formerly Exponent) publishes a public search API over its indexed pages; the guide is at https://www.aced.io/agents/agent-access.

Source: https://www.aced.io/blog/product-execution-interview-disney-plus
Source observed: 2026-11-05T01:49:32.832Z
Last checked: 2026-11-05T01:49:32.832+00:00
Indexed: 2026-11-07T03:38:20.319+00:00

---

# Product Execution Interview Example: Increase Disney+ Retention

![Aced Team](https://s3-us-west-2.amazonaws.com/pmlesson/2019/11/exponent-logo-fb.jpg)

Aced (formerly Exponent) publishes this mock product-execution interview for product-management candidates. George Perantatos, Product Director at Redfin, answers the question “Increase customer retention for Disney+ or Netflix,” with Chris Wilson, Senior+ Product Manager, conducting the interview. George’s distinguishing strength is judgment about when experimentation is useful and when qualitative learning, prototypes, or product judgment are more appropriate.

[View the related product-execution interview question](https://www.aced.io/questions?page=1&role=pm&type=execution&ref=exponent-blog.ghost.io). [George Perantatos](https://www.linkedin.com/in/georgeperantatos/?ref=exponent-blog.ghost.io) and [Chris Wilson](https://www.linkedin.com/in/automateiq/?ref=exponent-blog.ghost.io) are the interview participants.

## When to Experiment

Experimentation is not the default answer to every product decision. Choose experiments based on the effort, uncertainty, risk, and type of learning required.

### Use experiments when

- A change is fast and easy to test, such as button copy, colors, or positioning, and an experiment can reduce subjective debate.
- You need to estimate the magnitude of an effect rather than obtain only yes/no validation.

### Prefer other methods when

- The work is exploratory and early-stage, such as a new AI chat feature.
- Qualitative learning matters more than statistical significance.
- The change is low-risk and straightforward to measure without controlled testing.

For an early, speculative AI-based chat experience, talking with users and learning what they do can be more valuable than immediately measuring statistical outcomes.

**Takeaway:** Before designing an experiment, decide whether the problem is actually an experimentation problem or whether research, prototyping, qualitative feedback, or product judgment is the better tool.

## The Biggest Mistakes

Teams often jump to solutions and use experiments without first grounding themselves in the user problem. Two failure modes follow.

### Random testing without purpose

Teams may generate ideas in planning meetings, move elements around, change colors, or try different copy without solving a specific user problem. When results are mixed, they cannot determine what success means because they never articulated it.

### Paralysis when data is ambiguous

One metric may rise while another falls. Product managers should combine data with intuition and judgment rather than treating ambiguous results as an instruction to stop making decisions.

**Takeaway:** Experiments are inputs to decisions, not the decisions themselves. Start with the user problem and use judgment when results conflict.

## Example: Improve Disney+ Retention

- **Prompt:** Design an experiment to increase retention at Disney+.
- **Hypothesis:** A more personalized **Recommended For You** section will keep users subscribed longer.

The exercise concerns both a business problem—retention—and a user problem: people may not find content that feels relevant, may feel the service is not for them, and may not return. The recommendation section is only a possible solution after the underlying failure is understood.

### Validate Before Testing

Before defining control groups and metrics, establish the current state:

- What is the engagement rate on existing movie and TV recommendations?
- Where are users struggling?
- Is the issue content discovery, recommendation quality, discoverability, user interface design, or something else?
- Is **Recommended For You** visible but ignored, or under-exposed despite being popular when users see it?

A working assumption in the interview is that **Recommended For You** is highly visible—it appears on every first-time load—but its recommendations feel outdated or irrelevant. State assumptions explicitly so the discussion can be redirected if an assumption is wrong.

**Takeaway:** Work backward from the business problem to the user problem and then to the specific failure point. Do not accept a proposed hypothesis without first understanding what is broken.

## One Change at a Time

Possible explanations include poor recommendation quality, poor discoverability, and a user-interface problem. George selects one hypothesis: recommendation quality.

His specific example is to add watch-duration data to the recommendation model. Instead of using only which shows users watch, the model would also use how long they watch them. That is one input change and one model improvement.

Avoid combining multiple model signals, a UI change, and other changes in one variant. If several things change at once, it becomes unclear what caused the result. Isolate the most promising, highest-return change, learn from it, and then stack subsequent learnings.

**Takeaway:** Test one hypothesis at a time. Choose the smallest and most valuable learning you can isolate. If it fails, take a different route; if it works, build on it.

## Input Metrics vs. Output Metrics

Retention, weekly active users, and subscription churn are output or lagging metrics. For a change to **Recommended For You**, first measure the behaviors expected to cause the outcome:

- Are users clicking **Recommended For You**?
- Are they scrolling through it?
- Are they viewing recommended content?

These UI engagement metrics should move sooner than retention. Engagement changes may appear within days, while retention changes can take weeks or months. If the expected engagement does not improve, there is little reason to wait for retention data to confirm that the change is ineffective. The approach also makes the theory of change explicit: improved interaction with recommendations should correlate with more daily active users and lower churn.

**Takeaway:** Measure the proximate behavior change expected to produce the desired business result, while still tracking the eventual output metric.

## MVP Experiments

A small experiment can produce learnings faster than one large experiment, but highly uncertain ideas should be validated before engineering work and controlled testing.

- **Low risk, low uncertainty:** Proceed directly to a focused test.
- **High risk or high uncertainty:** Use prototypes and qualitative user research first to determine whether the idea feels better and is directionally sound.

It can be cheaper to test a prototype with users than to build and ship an experiment only to discover that the concept was far from useful.

**Takeaway:** Keep the first test small and focused; validate novel or uncertain ideas qualitatively before building them.

## Interviewer’s Perspective

Chris Wilson identifies three qualities that made George stand out:

- **Conviction about experimentation:** Not every organization runs experiments or wants to. A product manager should have a point of view rather than simply follow an expert’s instruction or act as an order taker.
- **Problem focus:** Candidates often discuss how to validate a solution before validating whether the problem exists. Strong candidates stay more committed to understanding the problem than to defending a preferred outcome.
- **Initiative:** George made assumptions and moved forward rather than asking unnecessary questions. A good product manager drives the conversation while remaining open to correction.

**Takeaway:** Senior and higher-level interviews evaluate judgment about when experiments belong in product development, not only the mechanics of executing them.

## What Could Be Improved

George says he could have been more structured in the moment. A clearer opening would have signposted the sequence: first discuss the user problem, then potential solutions, and finally how to measure them.

Signposting helps the interviewer follow the reasoning, redirect if the candidate is taking the wrong path, prepare follow-up questions, and evaluate the response. Chris adds that pausing and writing down the structure demonstrates deliberate thinking and gives the interviewer time to consider the next question.

**Takeaway:** In scenario questions, outline the structure before diving in—for example, “I’m going to cover X, then Y, then Z.”

## Senior+ Candidate Tips

Technical concepts such as statistical significance and sample-size calculations are table stakes. Higher-level skills include:

- Judging when to use experimentation tools.
- Deciding when to experiment versus ship.
- Deciding when to measure versus learn qualitatively.
- Knowing when to wait for data versus trust conviction.
- Deciding which experiments belong on the roadmap.

Even in a detailed hypothesis or dashboard question, explain whether the experiment should run at all and how it fits into product development. That strategic perspective can distinguish otherwise technically comparable candidates.

## Practical Takeaways

### Before the interview

- Develop a clear point of view on when experiments are useful and when they are not.
- Prepare specific examples of experiments that worked, failed, or should not have been run.
- Practice explaining your product philosophy, not only your process.

### During setup questions

- Lead with your framework for thinking about experimentation.
- Ground answers in specific examples.
- Challenge assumptions and show nuance.

### During scenario questions

- Outline your structure before diving in.
- Start with the user problem, not the solution.
- Make assumptions explicit and invite redirection.
- Work systematically while driving the conversation.

### When discussing experiment design

- Explain what you would research before designing the test.
- Choose one hypothesis rather than several.
- Start with input metrics that signal the expected behavior change.
- Keep the first test small and focused on the highest-return change.

Aced provides tech interview preparation, including free interview guides, insider tips, and courses. [Create your free account](https://www.aced.io/signup?src=blog).

---

[Read this page’s summary](https://www.aced.io/agents/api/public/v1/page-tldr?url=https%3A%2F%2Fwww.aced.io%2Fblog%2Fproduct-execution-interview-disney-plus&trace=t_679ff9ab98f3)

## Ask a follow-up about this page

Need a more specific answer? Ask the exact question you have. It returns a fresh, cited answer from Aced (formerly Exponent)’s knowledge base. No account or API key required.

Open this URL with your own question, URL-encoded. Include your question in q. All research inputs are optional; no account or setup.

`GET https://www.aced.io/agents/agent/faq?q={question}&trace=t_679ff9ab98f3`

[Tool guide and examples](https://www.aced.io/agents/agent-access) · [OpenAPI](https://www.aced.io/agents/api/public/v1/openapi.json) · [Page TLDR form](https://www.aced.io/agents/agent-tools/page-tldr?url=https%3A%2F%2Fwww.aced.io%2Fblog%2Fproduct-execution-interview-disney-plus). Opening a form does not execute a tool.