Skip to content

Growth

During interviews, how to navigate trade-offs in pursuit of growth
(Want advice on what interviewers are looking for? 👉 👈 )
Search
The arc of this question
1
Step 1: set context. State your product mission, business goal, and highest level metric. Describe who your product's different user types are and why they use the product today.
Step 2: go broad. Return to the growth question and propose a few different approaches, and discuss pros and cons.
Step 3: converge. Select one of your approaches. Describe mitigating factors and learnings or new information that could change your approach.
No results from filter

Growth is perhaps the widest ranging type of data interview question. Some versions of it are a more standard “Product Sense” interview, other flavors are focused on marketing and funnel optimization. Another variation centers on how to improve data-heavy features like algorithms. All of these get at analytical thinking and showing personal experience when trying to move numbers.
Practice question: how would you optimize Amazon shopping cart conversions?
This question is designed to get into talking about user segments and scenarios. It can often include a funnel optimization component and a blue sky brainstorming period. The interviewer is looking for answers that have nuance and reflect the varied usage patterns, scenarios and user types product has. Metrics can help demonstrate what those scenarios and users are like, prioritize approaches, and verify the approaches are working.
Search
Overall Score
Answer
Interviewer take away
1
1 - strong no 😡
Run some A/B tests on the color of the checkout button in the shopping cart, and on the call to action text in the button.
Skips Section 1 and Section 2 discussions completely. While these are fine minor optimizations to try, this answer fails to set a broader context around how much these might matter, what else should be considered, and what scenarios we should target among many possible scenarios.
@📓 Demonstrates experience
is also weak, since ignores other obvious levers like notifications, discounts, etc.
2
2 - no 🙁
The metric I would definitely optimize around is “resolution from cart” - which could mean purchase conversion or could mean deletion from cart. I would focus on getting users to take an action on their cart, using all tools, platforms, and surfaces we have available. We could send mobile notifications, emails, put toasts on top of pages, and have reminders about “did you want to [remove] or [buy it now]” with quick action buttons across the site. Having both actions in all these places - to either buy or remove - will make the aggressive prompting hopefully feel more like a reasonable triage than an upsell.
This answer mostly skips Stage 1 (goal of the shopping cart) and Stage 2 (brainstorming different options and discussing trade-offs). It reframes the question to an adjacent goal, but should have acknowledged the downsides of choosing this metric and other options considered.
@🦅 Changes altitudes
well from principle to examples. But it doesn’t convey a sense of what the single most important thing to do is and why.
@💬 Communication
style is overconfident at this stage.
3
3 - yes 🙂
Before trying to optimize shopping cart conversions, I would try to deeply understand what’s going on today. How does shopping cart abandon rate compare on mobile, desktop, tablet and voice devices like Alexa? We might want very different strategies based on each platform’s capabilities, like mobile notifications on mobile, for example. How does conversion rate today vary by country and user demographic? How does it vary by kind of product purchase: are very expensive items held in the cart a long time more like a “wishlist”? Does it seem like people forget to hit checkout on small items? Depending on these answers, our strategy might be really different.
@🤔 Asks good questions
before diving in.
@🦅 Changes altitudes
seamlessly from principle to concrete examples.
@💭 Structured thinking
shows a few key axes and illustrates how those might shape the answer significantly. This answer also shows the candidate is trying to make this more of a dialogue/conversation, and will get more information on what the interviewer is thinking and looking for, which is a strong
@💬 Communication
technique. This partial answer focuses squarely on Stage 1, setting context.
4
4 - strong yes 😍
The mission of the shopping cart is to carry promising items around with you to purchase when ready. While optimizing the shopping cart conversions is a powerful funnel that can directly lead to more revenue, it’s worth putting in context: Amazon also has “buy it now” that bypasses shopping carts altogether, and subscription purchases that setup a schedule for ongoing purchases. Finally, there’s also a Wishlist primitive and sharable Lists, which are even longer running lower intent research tasks than a Shopping Cart. Given this landscape, we should think about the Shopping Cart is high intent but not certain purchases, and our opportunity is to give a buyer more information or incentives to make a decision. Does all that make sense? (If so we can move forward to proposing some optimizations.)
Establishes strong principles to derive further decisions from. Uses
@💭 Structured thinking
to describe a spectrum of purchasing behaviors, with shopping cart in the middle. Nice
@💬 Communication
: checks for feedback before brainstorming, to make sure this foundation is reasonable.
No results from filter

We discover a common delay from shopping cart to purchase is from family members consulting each other before checkout - how might Amazon accelerate that?
This is a Step 2 follow-up question under the Amazon Shopping Cart prompt. It’s looking for an answer that brainstorms broadly and uses metrics and/or user segmentations to prioritize.
Search
Overall Score
Answer
Interviewer take away
1
1 - strong no 😡
Since families are often together in real life, I would focus on the Alexa in-home voice angle. Alexa can tell one family member that another needs their advice to close out a purchase.
This answer is really tough to engage with. While there’s a kernel of something promising, around focusing on an in-home scenario, there’s no sizing of how big an opportunity that is (how many families have Alexa devices?). Moreover, it’s not clear how the scenario would work: What is Alexa saying to whom? Why is that better than one person in the house talking to another? What other options were considered and why is this one the top choice? Missing
@💭 Structured thinking
,
@🤔 Asks good questions
and strong
@💬 Communication
.
2
2 - no 🙁
Since we’re focused on a collaborative family use case, I’d consider approaches that bring more social interaction into the product itself to reduce roundtrip time and bring more Amazon context into the conversation. For example 1) Amazon could launch a chat product sort of like Facebook Messenger. It could be a chat mole and allow you to create groups with other Amazon users/family members to discuss products. In this context, you can approve/reject ideas together, or propose alternatives; 2) Build a review queue into the product for parents to approve kids requests. This likely has some overlap with some business scenarios (team member needing boss approval for a purchase). Between these I would pick the chat product.
The initial social framework argument is a reasonable
@🤔 Asks good questions
, but the content that follows has too much implicit. For example, what exactly is the Amazon messenger and why is it better because it’s in context? Why would your whole family move over to it from your existing social discussion context? For the review queue, this assumes kids have their own Amazon accounts today - is that assumption reasonable? Raising these assumptions explicitly would have been better
@💬 Communication
and stronger
@💭 Structured thinking
. The choice of the messaging direction is never connected back to the original goal.
3
3 - yes 🙂
Since this insight is about family members, I will assume the same people are making joint purchases together repeatedly. As such, I’d err towards changes that may have a little setup and then are fast and powerful once configured. A few ideas: 1) a shared umbrella account concept (like Netflix) so the whole family can exist under one account for shared billing and private commenting. 2) Push recurring subscriptions inventory and item raking to make more decisions happen once and then be automatic. 3) Set joint goals and budgets in Amazon directly, so there’s a shared work area for bigger ticket items. Between these, I’d focus on a new umbrella account construct. It both gives more capabilities for purchase approvals for the kids use cases, opens up a new model of native social interaction (within umbrella account commenting/sharing), and allows for new ML ranking refinements based on family preferences. Said another way, it seems like the most systemic change that opens the door to many other optimizations.
While this user doesn’t breakup the family user scenarios into more granularity, it does offer
@💭 Structured thinking
for its brainstorm: medium-to-high setup, low incremental work, high return. It
@🦅 Changes altitudes
to several promising concrete ideas, and picks one with a reasonable justification. While this is a good answer, it does beg more follow-up questions what magnitude of improvements do you think may be possible with each of these examples? Given it is a medium effort setup, what mix of carrots or sticks do you think will be necessary to get families onto this new behavior pattern?
4
4 - strong yes 😍
There’s a range of family scenarios: 1) a couple making a decision together, 2) a parent making a decision with their kid, 3) a kid wanting parent approval to make a purchase. There’s likely a trade-off between frequency of purchasing and purchase amounts. For example, I’d guess the least frequent purchase types have the most transaction dollars going through it (eg: buy a new bicycle, fridge, etc) and that’s precisely why you want another family member’s opinion. So the question becomes, should we primarily optimize the number of purchases or dollars spent? As a simplifying assumption, let’s say dollars spent since that’s closely connected to our bottom line.
Let’s brainstorm improvements by purchase funnel stage. At the browse stage, Amazon could launch a surface like Pinterest to collect and discuss ideas privately within a family. At the comparison stage, we could create a surface for building your own comparison matrixes for competing products and filling them in collaboratively. At the shopping cart stage, we could let you 1-click share your cart via a link, and have recipients comment and up/down vote on items. At the final purchase stage, we could let account holders create rules for blocking any purchase > $X on a text confirmation (for the parent needing to approve child purchases on their account).
In a second we can dig into which of these ideas I’d prioritize. First, any questions so far?
Often great answers are more conversational, with clarifications or checkpointing for quick feedback along the way. While this answer was a longer monologue (it was its own mini Step 1-2), it hit the key points. It outlined some user personas showing
@💭 Structured thinking
, pointed out a subtle difference in goals (
@🤔 Asks good questions
) and took a clear stand on one to operate with. It had a framework for how to think about/categorize ideas (by funnel stage), and
@🦅 Changes altitudes
to propose a number of clear concrete options. It culled the list into a final proposal with a strong justification.
No results from filter
Practice, rinse, repeat
👉 When you’re ready, let’s look at all .
Below are more example Growth questions to work through.
Copy this doc
and use it as a worksheet and checklist!
Growth Questions Worksheets
5
Search
How would you optimize Amazon shopping cart conversions?
Set context: Go broad: Converge:
You’re the PM in charge of reactions for Facebook. How would you think about non-like reactions per post? How would this differ by reaction? Why/why not?
Set context: Go broad: Converge:
How would you roll out an algorithm improvement for driver matching at Uber?
Set context: Go broad: Converge:
“A” or “B” option—how do you know what to show to which communities of users?
Set context: Go broad: Converge:
How would you increase engagement?
Set context: Go broad: Converge:
How would you acquire more users?
Set context: Go broad: Converge:
Draw and explain company X’s marketing funnel
Set context: Go broad: Converge:
What would you do to optimize our funnel?
Set context: Go broad: Converge:
How would you leverage YouTube in order to reach our target audience?
Set context: Go broad: Converge:
How do you prioritize your experiments?
Set context: Go broad: Converge:
What do you do to understand behavior of your target customers?
Set context: Go broad: Converge:
How do we know if our signup page is “good”?
Set context: Go broad: Converge:
Why is it important if an A/B test result is statistically significant?
Set context: Go broad: Converge:
You run a bank - how would you decrease your fraudulent purchases rate?
Set context: Go broad: Converge:
What analysis would you use to understand if we should increase the price of an UberEats delivery?
Set context: Go broad: Converge:
How would you prioritize the different things you want to work on?
Set context: Go broad: Converge:
How would you build YouTube’s related videos engine?
Set context: Go broad: Converge:
How would you know whether to trigger a video results unit in Google Search?
Set context: Go broad: Converge:
How would you detect email spam?
Set context: Go broad: Converge:
How would you decide what five thumbnails to show for each YouTube video?
Set context: Go broad: Converge:
You’re a PM at Yelp. Create a system that recommends new restaurants to try.
Set context: Go broad: Converge:
You’re building a grocery store. How many checkout aisles do you want? What split between automated and cashier-run aisles do you want? How would you decide?
Set context: Go broad: Converge:
How much should Facebook be willing to spend per user acquisition?
Set context: Go broad: Converge:
You’re the PM on a Shopping website and your team has built a box that shows “Related Products.”You’ve run a 2-week experiment for this. How would you decide whether to launch?
Set context: Go broad: Converge:
If you’re Android and LG wants to build their own app store, how much would you be willing to pay them to not do that?
Set context: Go broad: Converge:
How would you improve YouTube for emerging markets?
Set context: Go broad: Converge:
How would you improve the Facebook photo experience?
Set context: Go broad: Converge:
How would you redesign Facebook Pages?
Set context: Go broad: Converge:
How would you prioritize showing “people you may know” vs showing an ad on Facebook?
Set context: Go broad: Converge:
How would you go about growing the Twitch streamer user base 10% YOY?
Set context: Go broad: Converge:
How do you prioritize competing features?
Set context: Go broad: Converge:
You’re the PM in charge of pricing. VP wants to lower from $70 to $69/year. Making your own assumptions develop the financial projections of the decision.
Set context: Go broad: Converge:
We're about to expand uberPool. What cities should we expand to?
Set context: Go broad: Converge:
Tell me about a time when you knew that something was wrong with a product and you had to convince a lot of other people.
Set context: Go broad: Converge:
Tell me about a product problem you fixed.
Set context: Go broad: Converge:
Question
How would you optimize Amazon shopping cart conversions?
Practiced
Type
🌱 Growth
Added by
DK
David Kossnick
Practice answer
Set context:
Go broad:
Converge:
Overall Score
💭 Structured thinking
🤔 Asks eigenquestions
🦅 Changes altitudes
📓 Demonstrates experience
💬 Communication
Feedback to self
Show hidden columns
Have questions that should be added?
Want to print your doc?
This is not the way.
Try clicking the ⋯ next to your doc name or using a keyboard shortcut (
CtrlP
) instead.