Product

Execution

Strategy

PM Interview

Strategy Guide
HOW TO CRACK ANY PM

INTERVIEW QUESTION

Behavioral

Technical
 Execution Questions
Metrics ​ 💯
These questions are popular at growth-oriented companies like Facebook. They ask you to pick
the right metric or metrics to gauge the success of a product and navigate different tradeoffs.

These questions are also common follow-up questions to product design questions. “You’ve
designed a cool product,” the interviewer says, “now tell me how you’d measure if it’s
successful.”

Sample questions​ 🤔
● What metrics would you choose to evaluate the success of Facebook events?
● Spotify is considering removing ads on its free tier. How would you measure if this is
successful?
● Duolingo is launching pilots of two new language courses. How would you judge which
one is more successful?
● On the top right of your LinkedIn feed, should you show "Sponsored Ads" or "People you
may know"?

Strategy​ ♟

It might seem like the obvious answer is “monthly active users” (MAUs) for consumer products
and “revenue” for enterprise products. But that’s too simple, and everyone will give that answer.

You can’t just make up metrics out of thin air. The key to answering these questions is to explain
the ​why​: why is the company doing this, and how could this help the company’s high-level
goals?

Please do not share this document with anyone as it is proprietary content with a unique
serial code in the bottom left to track piracy. The sharing, reproduction, resale or distribution
of this copyrighted work is strictly illegal and punishable by fines up to $250,000.
© Product Alliance

DF8S-PF27-FG9U
1. Know the north stars ​ 🌟
Your first order of business should be to understand the company’s big-picture goals. Revenue
and profit are obviously the company’s most important metrics, but they’re hard to measure, and
business teams often won’t divulge that information.

Instead, you want to look at the metrics that most directly correlate with revenue. There are five
types of tech business models, each of which has two key “north star” metrics:

In 4 of the 5 cases, revenue is roughly equal to MAU times some other quantity, so those two
metrics matter the most (Pricing-related metrics like ARPU/ARPA and the price of an ad are also
important, but usually out of a PM’s control). Check out the ​Product Management’s Sacred
Seven​ part on Data Science for more.

After asking clarifying questions, categorize the company’s business model and name the two
north star metrics.

2. Map it ​ 🗺
During the interview you should be connecting the feature/product north star metrics to the
company’s mission statement (which explains why the problem being solved is societally

Please do not share this document with anyone as it is proprietary content with a unique
serial code in the bottom left to track piracy. The sharing, reproduction, resale or distribution
of this copyrighted work is strictly illegal and punishable by fines up to $250,000.
© Product Alliance

DF8S-PF27-FG9U
important). For instance, if the proposal is for Dropbox to increase the free tier’s storage limits,
your mindmap could start off like this:

Then you want to brainstorm intermediate metrics changes that could show you’re making
progress toward one or more of the goals. Draw arrows to connect the feature to the
intermediate metric to the goal. For instance, increasing the free tier’s storage limit could
increase the number of shared folders. This could drive more businesses to upgrade (since
they’d be more reliant on Dropbox). As a result, this would help achieve Dropbox’s mission of
“building the world’s first smart workplace,” which presumably includes lots of digital sharing.
Talk through this logic with the interviewer and then add it to the mindmap, like so:

Please do not share this document with anyone as it is proprietary content with a unique
serial code in the bottom left to track piracy. The sharing, reproduction, resale or distribution
of this copyrighted work is strictly illegal and punishable by fines up to $250,000.
© Product Alliance

DF8S-PF27-FG9U
Brainstorm 5-7 intermediate metrics in this way. These will be the metrics you use to evaluate
the product or feature’s success. From here on out, the question gets a lot easier.

3. Prioritize & set goals ​ ✅

Once you’ve brainstormed a set of candidate metrics, write them all in a list ranked by
importance. If you could only use one metric to measure the success of this product, what would
it be? What if you had two? Three?

Choose the top three metrics and set numerical goals for each (this is probably going to be the
interviewer’s first follow-up question, so you’ll get bonus points for bringing it up voluntarily).
Summarizing these in a table on the whiteboard can be helpful.

For instance, if one key metric for the feature is to increase the number of files stored, the goal
could be to increase this number 30% in one year. Be ambitious — the kinds of growth-focused
companies that ask these questions love seeing candidates propose bold growth numbers.

4. A/B test ​ 📊
If your question involves a new feature or product, be sure to mention A/B testing at the end of
the interview. Explain that you’d set up experiments to test the new feature or product, measure
the three key metrics, and see if they hit the goals you outlined above. Growth-focused
companies love A/B tests as a way to prove out new ideas.

Most tech companies run 1% experiments (i.e. roll out the feature to a random 1% of users), but
that may not always work. If the feature is social, it’s quite useless if a random 1% of users have
it — people with the feature will have no friends to use it with — so it’s useful to give the feature
to everyone in a small, homogenous country like Portugal. Also note that some features, such as
increased Dropbox storage space, can’t be easily taken back, so you have to be willing to live
with the feature forever even if it doesn’t test well.

If the question is about finding metrics for an existing feature, you can propose some A/B tests
to grow the key metrics you laid out earlier. For instance, if you were asked to name the key
metrics for Facebook birthdays, you could have said that a key metric is the number of
birthday-related messages sent, and to grow that you could A/B test a feature that encourages
users to send pre-filled birthday messages to friends. This veers into product design, but if you
can think of some simple feature ideas quickly here, you can earn massive bonus points.

Please do not share this document with anyone as it is proprietary content with a unique
serial code in the bottom left to track piracy. The sharing, reproduction, resale or distribution
of this copyrighted work is strictly illegal and punishable by fines up to $250,000.
© Product Alliance

DF8S-PF27-FG9U
Rubric​ 📄
Score yourself from 1-5 on the following criteria:

● Strategic understanding:​ Did you understand the key metrics and mission for the
company? Could you tie this product’s success to the company’s success, as with the
mind map?
● Creativity: ​Did you come up with several interesting metrics beyond the obvious MAUs
and DAUs?
● Prioritization: ​Did you rank the most important metrics for your feature and explain why?
● Love of testing: D ​ id you set ambitious but realistic metrics goals and explain how you’d
test the product to see if it’ll hit those goals?

Please do not share this document with anyone as it is proprietary content with a unique
serial code in the bottom left to track piracy. The sharing, reproduction, resale or distribution
of this copyrighted work is strictly illegal and punishable by fines up to $250,000.
© Product Alliance

DF8S-PF27-FG9U
Analytical​ ⚗
​

These questions are less common but pretty hard. They involve “root-causing” a problem that
you identified in the metrics for a hypothetical product.

Sample questions​ 🤔
It’s easiest to understand these questions by seeing examples:

● YouTube watch time went down 5%. How would you diagnose what happened?
● Uber rides decreased 10%. What do you do?

Strategy​ ♟

These questions test whether you can get to the heart of the problem. The exact change in
metrics doesn’t matter that much, and it’s not too hard to think of a solution once you’ve found
the root cause of a problem. The strategy for these questions, then, is to first narrow down the
(usually vague) question to a very specific problem, and second to rule out possible causes until
you find the truth.

1. Sharpen the problem ​ 🔻

Part one — finding the core problem — involves a back-and-forth dance of questions. Usually
the problem is vague — “this metric went down X%” — but the interviewer actually has a more
specific sub-problem in mind. For instance, a problem phrased as “YouTube watch time went
down 5%” might actually be “YouTube watch time for kids in Australia on the iOS app went down
5% last week.”

To narrow down the problem, you want to ask the following type of question repeatedly: “did the
problem only occur for particular ___?” The blanks could be filled with:

● User groups (age, user segment, usage levels, etc.)

● Time periods
● Countries / geographical areas
● Operating systems or browsers

Please do not share this document with anyone as it is proprietary content with a unique
serial code in the bottom left to track piracy. The sharing, reproduction, resale or distribution
of this copyrighted work is strictly illegal and punishable by fines up to $250,000.
© Product Alliance

DF8S-PF27-FG9U
The back-and-forth might go like this:

● Interviewer: “Uber rides decreased 10%. What do you do?”

● You: “Let me first narrow down the problem. Did this decrease happen all across the
world, or only in a particular country?”
● I: “Just in Germany.”
● Y: “When did we notice this change? Has it been consistent?”
● I: “We noticed it last Saturday, and it’s been happening consistently every day since then.”
● Y: “Has it been concentrated in particular times of the day?”
● I: “Mostly in the late evenings or early mornings.”
● Y: “Did this happen for only certain kinds of users?”
● I: “It affected our business travelers more than personal travelers.”
● Y: “Did it happen more on Android vs. iOS?”
● I: “No, the decrease was happening consistently.”
● Y: “OK, so I’ve refined the problem to: Uber rides in Germany decreased by 10% among
business travelers starting last Saturday, chiefly in the late evenings and early mornings.”

2. Root-cause ​ 🔍
Once you’ve clarified the problem, you can then figure out the root cause of the problem. You’ll
want to consider both internal causes (your product change broke something) and external
causes (a change in the rest of the world broke something).

Start by investigating external causes. Assume for now that your product didn’t change at all.
Could any changes in the rest of the world — cultural shifts, new competitors, new laws, etc.
— have caused the problems you’re seeing? Examine your precise problem phrasing and spend
a few minutes writing down possible external causes that fit the data you see. Some examples of
external root causes:

● In the Uber example, it’s possible that the problem was caused by Germany banning Uber
at airports. This would explain why business travelers were affected more and why the
problem was more concentrated late at night, when public transit isn’t running.
○ Knowing the speed at which the metric changed matters too. If the change was
gradual, it’s probably due to some cultural change. If it was sharp, it’s probably due
to some sudden introduction of a new law or competitor.
● Suppose Square’s usage among small merchants in Boston slowly decreased from
September to November, but the same pattern didn’t hold in Atlanta. It could be that
Boston’s winter started earlier than usual, thus forcing farmers’ markets (who use Square
a lot) to close shop earlier than expected.
○ As this example shows, it’s useful to get counterfactual data too — where did the
problem n​ ot​ happen?

Please do not share this document with anyone as it is proprietary content with a unique
serial code in the bottom left to track piracy. The sharing, reproduction, resale or distribution
of this copyrighted work is strictly illegal and punishable by fines up to $250,000.
© Product Alliance

DF8S-PF27-FG9U
 ● Suppose Airbnb’s average revenue per booking decreased all summer in southern
California. This could have been caused by many budget travelers joining the platform
(say, if many airlines started discounting flights to Los Angeles) or many wealthy travelers
leaving (say, if luxury hotels introduced new loyalty programs). To clarify, you’d probably
want additional data on the month-over-month changes in, say, under-$100 bookings
and above-$500 bookings.
○ The actions of complementary or competing businesses are common external
causes for metrics changes.
○ This example also shows how one metric change can have two or more
polar-opposite potential causes. This is especially true with averages (as in this
example) or metrics described as a fraction (where either the numerator or
denominator could have changed).
● Suppose Tinder DAUs dropped sharply overnight, but only among college-educated users
in big cities. One likely explanation is that a new dating app that caters to young urban
professionals just launched.
○ It’s often hard to tell the difference between cultural and business shifts;
college-educated users could also have left because of a Tinder brand issue. But
cultural changes happen gradually, while business changes happen sharply, so use
the timeframe to your advantage (most interview candidates wouldn’t think about
this, so call out this logic in the interview).
○ Interviewers love asking analytical questions that mention “overnight” changes, for
some reason.
○ Some changes in people’s behavior are sudden, though. Retail sales spike
immediately on Thanksgiving, Uber usage dropped sharply as soon as COVID-19
sent the US into lockdown, dating app usage drops as soon as students leave
college campuses after finals, etc. Cultural changes in opinion, like something
becoming seen as old-fashioned, are slower.

If you’ve ruled out all external changes and know that the change wasn’t caused by any cultural,
legal, environmental, or business trends, you now know that the problem was caused by some
change or bug in your product — an internal change. Internal changes are probably to blame if
the metrics dropped on one browser or OS but not another, or if changes were identical across
demographics and geographies. If you suspect an internal change is to blame, ask if there have
been any recent code pushes or A/B tests in the affected areas.

Generally, we advise investigating external causes first and only moving on to internal causes if
all external causes have been ruled out. This is because most candidates don’t even think about
external causes, so you’ll look better by carefully considering them before jumping to
conclusions. You can brainstorm both kinds at once if you can’t think of many external causes,
though. Keep in mind that this depends on the situation – if you are investigating a metric that
changed after a product release, then you should look at internal metrics first. On the other

Please do not share this document with anyone as it is proprietary content with a unique
serial code in the bottom left to track piracy. The sharing, reproduction, resale or distribution
of this copyrighted work is strictly illegal and punishable by fines up to $250,000.
© Product Alliance

DF8S-PF27-FG9U
hand, if there is a random drop in metrics you want to investigate external factors (news articles,
competitors releasing products, etc.) before bothering your team.

We recommend organizing all potential root causes in a ​direction table​ and comparing them to
the data you’ve observed to rigorously test them all. For instance, if you found that YouTube
usage declined among kids but not adults, and in the US but not in Europe, you could make this
table to see if each hypothesis matches the observed data:

We recommend putting both plausible and less-plausible potential root causes in this table. Even
if you can rule out a root cause immediately, put it in the table to show you thought about it
logically.

If you get down to more than one plausible root cause, try asking more questions to get data
points that would only leave one root cause as plausible. It could be as simple as asking if a
particular event has happened (e.g. has a law like this been passed recently?) or as advanced as
digging into historical data (e.g. did a drop of a similar magnitude happen the last time a
competitor launched?)

Rubric​ 📄
Score yourself from 1-5 on the following criteria:

● Information gathering:​ Did you ask for more information before making an opinion? Did
you sharpen the problem all the way, or did you miss some nuance the interviewer was
hiding?

Please do not share this document with anyone as it is proprietary content with a unique
serial code in the bottom left to track piracy. The sharing, reproduction, resale or distribution
of this copyrighted work is strictly illegal and punishable by fines up to $250,000.
© Product Alliance

DF8S-PF27-FG9U
 ● Creativity: ​Did you think of any unusual potential causes, especially ones that fit the data
especially well?
● Logic: ​Does your root cause fully explain the metrics that changed? Does your cause not
affect the metrics that didn’t change?
● Technical skills:​ If the problem was internal, were you comfortable talking through
potential bug sources, configuration errors, etc.?

Please do not share this document with anyone as it is proprietary content with a unique
serial code in the bottom left to track piracy. The sharing, reproduction, resale or distribution
of this copyrighted work is strictly illegal and punishable by fines up to $250,000.
© Product Alliance

DF8S-PF27-FG9U
Crisis Response​ ​⚠

These questions ask how you respond to problems that arise while you’re leading the team to
deliver a product. Think of them as, “something went wrong — how do you handle it?”

Sample questions​ 🤔
These questions feel a bit like behavioral questions, but they’re more about hypothetical
situations:

● Your supplier has said a delivery will be delayed, so you can’t ship a hardware product in
time to hit a customer’s deadline. How do you respond?
● You work for Uber, and Spain is considering a law that would ban Uber in the country.
What do you do?
● You’re a PayPal PM and found that the app has been hit with a zero-day exploit, putting
user data and money at risk. What are your first steps?
● Facebook launched a new privacy settings page and got sharp criticism in the media
immediately. You’re the PM — what do you do?

Strategy​ ♟

As the sample questions show, these questions are all about operations and leadership. The
operations part requires knowledge of regulatory and legal compliance, as well as privacy,
security, and supply chain topics. We don’t have room to cover these detailed areas, so we’ll
point you toward the Law & Policy chapter of P​ roduct Management’s Sacred Seven​.

Once you have the operations knowledge down, the key to this question is showing that you’d
have a calm, logical, and structured approach to the crisis instead of panicking. We recommend
using this modified version of the tech industry’s ​postmortem framework​, which explains the
steps to responding to a crisis or failure:

1. Decide:​ work with internal teams to agree on a response strategy

2. Resolve:​ fix the immediate bug, exploit, or other problem as quickly as possible
3. Mitigate:​ reduce the damage of the impact, such as by helping affected customers
4. Prevent:​ put systems in place to ensure the problem doesn’t happen again

Please do not share this document with anyone as it is proprietary content with a unique
serial code in the bottom left to track piracy. The sharing, reproduction, resale or distribution
of this copyrighted work is strictly illegal and punishable by fines up to $250,000.
© Product Alliance

DF8S-PF27-FG9U
It’s important to do these steps in order, since they go from shortest-term to longest-term.
Earlier steps are more time-sensitive, but later steps are more important in the long run.

For instance, suppose the question was, “you’re a Lyft PM and found that your payment
processor failed, so you couldn’t pay your drivers today. What do you do?” Your response should
be to sketch out a table with each of the four postmortem steps:

The “decide” step is pretty boilerplate — get everyone internally to agree. It’s important to
mention nonetheless so you show the interviewer that you’re a team player and would respect
everyone’s opinions.

The difference between the “resolve” and “mitigate” steps is subtle but important. Resolving the
issue means returning the world to the pre-crisis state, including making affected customers
whole. This is usually a more technical step. Meanwhile, mitigation is more about PR for your
company and “damage control;” communication to the media and giving customers bonuses (on
top of whatever they lost from the crisis itself) are the main tools here.

Please do not share this document with anyone as it is proprietary content with a unique
serial code in the bottom left to track piracy. The sharing, reproduction, resale or distribution
of this copyrighted work is strictly illegal and punishable by fines up to $250,000.
© Product Alliance

DF8S-PF27-FG9U
Rubric​ 📄
Score yourself from 1-5 on the following criteria:

● Composure: ​did you have a calm, measured response to the crisis? Did you not
overreact?
● Teamwork: d ​ id you explain how you’d work cross-functionally to solve the problem? Did
you take other teams’ needs into account when developing the plan?
● Thoroughness: ​does your proposed resolution actually stamp out the problem? Does
your proposed prevention stop the problem from happening again?
● Speed: c
​ ould your proposed steps be done in a matter of days?

Please do not share this document with anyone as it is proprietary content with a unique
serial code in the bottom left to track piracy. The sharing, reproduction, resale or distribution
of this copyrighted work is strictly illegal and punishable by fines up to $250,000.
© Product Alliance

DF8S-PF27-FG9U
 Thank you for reading!
Visit www.productalliance.com for more
helpful resources.

Please do not share this document with anyone as it is proprietary content with a unique
serial code in the bottom left to track piracy. The sharing, reproduction, resale or distribution
of this copyrighted work is strictly illegal and punishable by fines up to $250,000.
© Product Alliance

DF8SPF27FG9U