Spark live coding assignment

Live coding Spark assignment that I did during one of my interviews

Coding assignment

Session Data - Holds aggregated information about a full session , a session is a visit to a website!

Path: /data/session_data.parquet

Columns:

tenant_uuid – uuid of the customer
session_ts: timestamp of the start of the given session, in ms
session_uuid: unique ID for the session in Glassbox
struggle_types: an array of dictionaries representing various struggles and their aggregated values throughout the session.
is_converted
struggle_converted

user_action - Holds information about specific “actions” (predefined list) performed during a session

Path: /data/user_events/*.parquet

Columns:

tenant_uuid – uuid of the customer
date
dom_element – element on which the action occurred
session_ts - timestamp of the start of the given session, in ms
session_uuid: unique ID for the session in Glassbox
potential_revenue: collected revenue
client_action: name of the action that triggered a row in this table (‘click’, ‘scroll’, ‘load’, ‘hover’)
struggle_score: struggle score for the session up to this action.

Take few minutes to learn and get acquintance with the data before proceeding further

Exercise 1

Compute for each and session, and for each tenant

Number of click actions in the session
Number of unique dom_elements in session
Number of clicks that has a potential revenue (more than 0) in the session
Number of page loads in the session

Exercise 2

A revenue is considered lost if the session is not converted. (is_converted=False) We would like to compute the daily average revenue loss by session, for each struggle.

Since revenue is saved in the action table, while the struggles as a dictionary in the session table, the computation for struggle revenue loss by session is:

We will define a struggle potential revenue as :

sum(potential_revenue) * (#struggle_type_occurences / #struggles in session)

example:

sergeioff / sparklivecodingassignment Goto Github PK