Analyze document collaboration patterns

You are given two CSV files and asked to analyze collaboration behavior on documents.

File 1: document_activity.csv

Each row is a document view event.

Schema:

• page_id STRING — unique document ID
• user_id STRING — user who viewed the document
• viewed_at TIMESTAMP — event timestamp in UTC
• is_creator BOOLEAN — true if the user is the document creator, false otherwise
• collaborator_source STRING NULL — for non-creators, how the user arrived at the document (for example: invite link, email invite, search, direct). This field may be NULL for creators.

File 2: users.csv

Schema:

• user_id STRING — unique user ID
• country STRING — user country

Assumptions

• Join key: document_activity.user_id = users.user_id .
• Use the full dataset; do not filter by date.
• A user may view the same document multiple times, so for collaboration questions you should deduplicate to distinct (page_id, user_id) pairs unless a task explicitly asks for event counts.
• A document may have one creator and multiple collaborators.

Tasks

1. Compute the distribution of collaborator acquisition sources.
   • Consider only non-creators ( is_creator = false ).
   • Define the metric on distinct collaborator-document pairs (page_id, user_id) .
   • Required output columns:
     • collaborator_source
     • collaborator_count
     • collaborator_share
2. Given an input user_id, find which other users collaborated with that user the most.
   • Two users are considered to have collaborated if they both appeared on the same page_id at least once.
   • Count distinct shared documents, not view events.
   • Exclude the target user themself.
   • Required output columns:
     • target_user_id
     • co_user_id
     • shared_page_count
   • Return the user or users with the maximum shared_page_count .
3. Find which country's users are most likely to collaborate with others.
   • To make the metric precise, define a user as having "collaborated with others" if they appear on at least one document with at least one other distinct user (that is, the document has 2 or more distinct users total).
   • For each country, compute:
     • collaborating_users = number of distinct users in that country who collaborated with at least one other user
     • total_users = number of distinct users in that country from users.csv
     • collaboration_rate = collaborating_users / total_users
   • Required output columns:
     • country
     • collaborating_users
     • total_users
     • collaboration_rate
   • Rank countries from highest to lowest collaboration rate.