Example question

How many times did individuals aged 20–30 in Berlin use Tinder yesterday?

Key concepts

• Grouping and aggregate functions.
• Joins (left, inner, outer, etc.).
• Appending data sources (union).
• Filtering data with complex conditions; understanding the difference between where and having.
• Window functions (lag and lead).
• De-duplicating, sorting, and managing missing data.

How to learn

• sqlzoo: Hands-on practice questions.
• sqlcourse: Breaks down concepts with examples and exercises.
• w3schools: Covers more topics but with simpler challenges.
• Mode: Visual code illustrations with examples and practice problems.

Practice consistently. Engage daily for at least two weeks leading up to your interviews.

Your journey in the data realm begins with SQL.

Example question

How many Uber rides took place in Los Angeles last week? What was the average revenue per ride?

Key concepts (Python)

Libraries - Data manipulation and analysis: Pandas, Numpy.

Dedicate 80% of your Python study to these areas. Master SQL concepts within Python, like groupby and join.

You’ll want to engrain this code into your memory:

import pandas as pd

• Visualization: Matplotlib, Seaborn.
• Machine Learning: scikit-learn.
• Statistics: statsmodels.

Variable types - list, tuple, dictionary, set, int, float, boolean, etc.

Features - Conditional statements: if, elif, else. - Loops. - Functions: def. - List manipulations: append, remove.

How to learn

• 10 minutes to pandas: Pandas official user guide.
• codecademy Learn Python 3: Covers important features with quizzes and projects.
• data analysis with python: Covers key libraries for data analysis.
• learnpython.org: Practice questions.

Example question

How would you assess the health of Facebook Groups?

Key concepts

Aligning Goals Understand how company objectives align with product aims. For instance, Meta’s mission — “Giving people the power to build community and bring the world closer together” — informs Instagram Stories, promoting frequent content creation and app usage.

Growth funnel: AARRR

Picture my Medium:

1. Acquisition: You discovered this article via Google.
2. Activation: You found it valuable and subscribed.
3. Retention: You continue to read my articles.
4. Revenue: You support me by buying me a coffee or hiring me for coaching or consulting.
5. Referral: You share it with friends.

Funnel Analysis Where do users drop off? What hurdles cause churn? For example, a lengthy sign-up process might deter users. What tactics could bring users back? Perhaps Thursday evening email notifications double the click rate compared to Monday mornings.

Evaluation Metrics 1. North star: The main indicator of product health. For Facebook, it’s Monthly Active Users (MAU). 2. Secondary: Supporting indicators like user retention rates and new user growth. 3. Guardrails: Business metrics to watch for negative impacts, such as the percentage of fake accounts on Facebook.

Troubleshooting Suppose Airbnb searches decline by 35% today. How would you identify the problem?

How to learn

• Cracking the PM interview: Learn to approach problems as a PM.
• Exponent PM mock interviews: Observe PMs from leading tech firms tackle product challenges in real-time — a priceless learning tool.
• Sequoia’s Product blog: Stay informed with insights from industry leaders.

Example question

Your PM wishes to launch YouTube Shorts. How would you assess the success of this feature?

Key concepts

• Understanding causation vs correlation.
• Evaluation metrics.
• Statistical concepts for A/B testing: T test, P value, Z score, Significance level, Confidence interval, Minimum detectable effect, Sample size estimation.
• Recognizing effects:
  1. Network effect (e.g., Bumble, Uber).
  2. Novelty effect when users are more responsive to newly launched features.

[Bonus] When A/B testing is not suitable 1. Small user base. 2. Brand-new app status. 3. Resource constraints.

Steps for Running an A/B Test

1. Clarify the objective: Gauge user reactions to YouTube Shorts.
2. Set metrics
   • Goal metric: YouTube daily active users.
   • Secondary metrics: Percentage of users engaging only with Shorts, overall time spent on the platform.
   • Guardrail: Time spent on traditional videos (where ad revenue is generated).
3. Form hypothesis: YouTube Shorts will boost daily active users.
4. Create experiment groups
   • Control: Users do not see YouTube Shorts (status quo).
   • Experiment: Users can see and interact with Shorts.
5. Calculate sample size
   • Choose significance level (alpha).
   • Determine the minimum detectable effect.
6. Consider potential complications
   • Should all creators have access to Shorts or only a select few?
   • Conduct a randomized user test or a location-based test to minimize network effects.
7. Decide on experiment length
   • It should run for at least a week to accommodate daily patterns.
   • It may need to extend longer with a smaller user group.
8. Perform significance test on goal metric & evaluate other metrics.
9. Make a launch decision.

How to learn

Udacity’s A/B Testing course by Google is my top recommendation. It covers everything mentioned here. Watch it twice: grasp the basics on the first viewing and delve deeper on the second. If you're short on time, read a course summary, but I highly recommend completing the full course and all exercises.

Example question

Can you illustrate the distribution of Tinder matches on a given day?

Key concepts

• Common distributions: Comprehend binomial and normal distributions.
• Central limit theorem: Understand how sample means form a normal distribution, regardless of the population distribution.
• Regression to the mean: Recognize that extreme values tend to revert towards the average.
• Probability and Bayesian probability: Differentiate between classical and Bayesian probability approaches.
• Variance, standard deviation: Measure data variability.
• Mean, median, percentiles: Calculate central tendencies and data positioning.

How to learn

• Probability & statistics from brilliant.org: Offers clear illustrations and practical problems to solidify your understanding.

Example question

How would you design a news feed ranking algorithm?

Key concepts

• Types of ML models:
  1. Regression: linear regression.
  2. Classification: logistic regression, decision trees.
  3. Clustering.
• Supervised vs. unsupervised learning.
• Evaluation metrics: accuracy, precision, recall, F score, AUC.
• Cross-validation.
• Bias-variance tradeoff: Find the right model complexity to avoid overfitting or underfitting.

How to learn

• Machine Learning with Python: This introductory course covers common models and their applications.
• geeksforgeeks: Covers a broader range of topics and advanced concepts.
• ML cheatsheet: A quick reference for each model and key statistical concepts.
• An Introduction to Statistical Learning: A textbook used in Columbia University’s Stats master’s program. If you have over six months to job hunt or aim for a career in Machine Learning, I highly recommend this book for its systematic and technical approach.

Specific visualization tools like Power BI or Tableau

Don’t worry about mastering tools like Power BI or Tableau unless your desired role specifically requires it. Tech companies often prefer Looker or Databricks for dashboarding, which you can quickly learn on the job. Tableau and Power BI are generally more relevant for data analysts.

Building Deep Learning models

Avoid diving deep into constructing complex models. While it’s tempting to engage in developing advanced technology, entry-level data science roles require a broad understanding of machine learning rather than deep specialization. Positions that require extensive deep learning skills typically have titles like “Machine Learning Engineer” or “Applied Scientist.”