Prarthana Insight Data Engineering Project

AdReportAnalytics for Ad Performance on Facebook

Prarthana Bhattarai

Insight Data Engineering

Motivation• Optimize expenditure on Facebook Ads

• Optimal level of bids More results at lower cost

• Weekly and monthly reports on ad performance

Image Source: Prweek.com

Data Pipeline

Batch Processing

Marketing API

Raw Data

Data ProcessKafka Message Sent to HDFS

Processed data stored in Cassandra

Challenges• Data Extraction • Managing Schema

Data Extraction

Image Source: Facebook.com

Managing Schema

Ad Set

Bid Info

Clicks

Targeting Specs

Ad Account

Active Ad Campaigns

Reach

ImpressionsAd SetAd Set

Next Steps• Integration of pipeline into Kuhcoon’s system• Cost: 100k customers at $23/day

Flask

KafkaHDFSSpark

Cassandra

Zookeeper

m3.large

t2.microKafkaHDFSSpark

Cassandra

KafkaHDFSSpark

Cassandra

KafkaHDFSSpark

Cassandra

Prarthana Bhattarai• BA, Mathematics, Mount Holyoke College• Resolution Fellow, 2014 -2015 • Loves to hike and read!