← Meta Interview Insights

Meta·Data Scientist·Technical Phone Screen·Intermediate

Intermediate
May 2026

Summary

SQL-heavy screen for a Data Scientist role at Meta, centered on hashtag follow behavior data. Nothing too wild but the percentage breakdown question had a bit of a wrinkle that I didn't fully anticipate going in.

Questions Asked (2)

Q1

Given a table of hashtag follow events and a table of hashtag safety classifications, which hashtag source (e.g. feed vs hashtag page) gained the most followers on a given day?

Product Analytics & MetricsData Modeling
Author's notes

Pretty straightforward aggregation.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clarifying the schema and definitions: what constitutes a 'follower' (e.g., a follow event), what are the possible hashtag sources, and how safety classifications might affect the analysis. Then write a SQL query that joins the follow events with the safety classifications, filters for the given day, groups by hashtag source, and counts distinct followers. Finally, consider edge cases like multiple follows by the same user and whether to include only safe hashtags.

Pro tip: Demonstrate awareness of data quality and business context: mention that you would check for duplicate follow events, bot activity, and whether safety classifications should filter out unsafe hashtags before ranking sources. Also, clarify if 'gained the most followers' means net new followers or total follows, as this can change the interpretation.

1. Clarify requirements and definitions

Ask clarifying questions to understand the tables, the definition of a follower, the time zone for 'day', and whether safety classifications should be used as a filter or just for segmentation.

2. Explore the data

Examine sample rows from both tables to understand columns, data types, and potential join keys. Check for missing values or anomalies.

3. Write the query

Construct a SQL query that joins the follow events with hashtag classifications, filters for the target day, groups by hashtag source, and counts distinct users (or follow events) per source.

4. Validate and handle edge cases

Check for duplicate follow events, users following multiple times, and whether to include only safe hashtags. Consider using window functions or subqueries to deduplicate if needed.

5. Interpret and present results

Rank the sources by follower count, identify the top source, and discuss any caveats or additional insights (e.g., safety classification impact).

Key Points to Mention

  • Definition of a follower: distinct user vs. follow event, and whether to count net new follows.
  • Join logic between follow events and safety classifications (e.g., on hashtag ID).
  • Filtering by date: using event timestamp and handling time zones.
  • Grouping by hashtag source and using COUNT(DISTINCT user_id) to avoid double-counting.
  • Handling safety classifications: whether to exclude unsafe hashtags or analyze them separately.
  • Data quality checks: duplicates, bots, and missing data.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q2

Of the users who followed a hashtag via the 'hashtag page' source, what percentage followed hashtags that are flagged as 'violating'?

Product Analytics & MetricsRoot Cause Analysis
Author's notes

This is where I fumbled a little.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Clarify the metric definition and data sources, then outline a SQL-based calculation that filters follows by source = 'hashtag page' and joins to a hashtag violation flag. Compute the percentage as the count of violating hashtag follows divided by total follows from that source, and discuss potential data quality or definitional issues.

Pro tip: Proactively mention that 'violating' hashtags may be flagged after the follow occurred, so you should consider using the flag status at the time of analysis or a time-based join to avoid temporal bias.

1. Clarify definitions and assumptions

Confirm what 'hashtag page' source means (e.g., a specific UI element or referral source) and how 'violating' is defined (e.g., policy violation flag). State assumptions about data availability and time windows.

2. Identify relevant tables and fields

Locate the follows table with source information, the hashtags table with violation flags, and any necessary join keys. Ensure you have timestamps for both follow events and violation status if needed.

3. Write the calculation query

Use SQL to filter follows where source = 'hashtag page', join to hashtags on hashtag_id, and compute the percentage of those follows where the hashtag is flagged as violating. Use COUNT or SUM with CASE statements.

4. Validate and interpret results

Check for data quality issues (e.g., missing flags, duplicates) and consider edge cases like hashtags flagged after the follow. Interpret the percentage in context of platform health and moderation.

Key Points to Mention

  • Define the numerator and denominator clearly: numerator = follows from hashtag page source where hashtag is violating; denominator = all follows from hashtag page source.
  • Consider temporal aspects: violation flags may change over time, so use the flag status as of the follow date or current status with caveats.
  • Ensure proper join between follows and hashtags tables, and handle potential duplicates or missing data.
  • Discuss data sources: likely a follows table with source column and a hashtags table with violation flag.
  • Mention potential segmentation: break down by time, user demographics, or hashtag category for deeper insights.
  • Highlight the importance of this metric for trust and safety, and suggest follow-up analyses if the percentage is high.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.