Slide 1: Snowflake at Sharethrough
Snowflake at Sharethrough Joseph Bates Director of Analytics
Slide 2: Sharethrough's Old World
Sharethrough’s Old World Minimum Viable Analytics No personnel, terrible performance, unacceptable limitations, no logical data model.
Slide 3: Why Snowflake?
SNOWFLAKE 2017-06-20 Why Snowflake? • One-stop shop • Performance, scale, flexibility • Ease of integration • Enablement of self-service • Quality of support
Slide 4: Jul 2015 Sep 2015 Nov 2015
SNOWFLAKE 2017-06-20 Jul 2015 Sep 2015 Nov 2015 Jan 2016 Mar 2016 May 2016
Slide 5: Amazon S3, Airflow, Snowflake Database, MicroStrategy
SNOWFLAKE 2017-06-20 Amazon S3 • System of Record • mix of JSON, AVRO,Parquet, and XML Airflow • ETL gatekeeper, integrity enforcer • base / agg fact generator Snowflake Database • “Dump everything here" data layer • ~140TB comp., or 10X the library of congress MicroStrategy • Defines and relates al
Slide 6: Fivetran EL
SNOWFLAKE 2017-06-20 Fivetran EL No transformations “Just works. (Even for Salesforce!)
Slide 7: Sharethrough's New World
Sharethrough’s New World … is a better world Upgrades from the top of the pipeline to the end user make for a happy business.
Slide 8: Snowflake Exposure
SNOWFLAKE 2017-06-20 Snowflake Exposure “This is a way for us to generate new analytics more quickly, by tapping an analytics-friendly DB to our clickstream.”
Slide 9: Snowflake Infection
SNOWFLAKE 2017-06-20 Snowflake Infection “I guess now that we have this level of aggregation against the logs, we could join it to our other source systems.”
Slide 10: Snowflake Epidemic
SNOWFLAKE 2017-06-20 Snowflake Epidemic “Is it data? Then dump it into Snowflake.”
Slide 11: Onward
Onward The perpetual lightbulb has been turned on Analytics is the lifeblood of the company, from the execs to AM to finance and beyond, and our short stack ensures we can keep pace.
Slide 12: EL(T) Snowflake ETL (Airflow)
SNOWFLAKE 2017-06-20 EL(T) Snowflake ETL (Airflow) Diagnostics Analytics BI Personnel
Slide 13: Base layer
SNOWFLAKE 2017-06-20 Base layer Maximize information, minimize volatility
Slide 14: Aggregation Layer
SNOWFLAKE 2017-06-20 Aggregation Layer Business as usual - alarms - ops tools - product launches - finance, exec metrics - OKRs and other KPIs
Slide 15: Self-service Analytics Adds Momentum
PRESENTATION SNOWFLAKE 2017-02-07 NAME - XX/XX Self-service Analytics Adds Momentum At every level of the enterprise From Data Science to Ops to Finances, removing barriers to data energizes the business, increases transparency, and emboldens decision-making.
Slide 16: Bullish on Snowflake
SNOWFLAKE 2017-06-20 Bullish on Snowflake Rapid support turnaround Market trends and “stickiness” Deployment pace New integrations Thought leadership in Cloud DWH Strength of core tech Transparency Value, and clarity of pricing
Slide 17: Questions / Comments
SNOWFLAKE 2017-06-20 Questions / Comments jbates@sharethrough.com https://www.linkedin.com/in/jcb4th
Slide 18: Appendix
SNOWFLAKE 2017-06-20 APPENDIX
Slide 19: Analytics Initiatives
SNOWFLAKE 2017-02-03 © SHARETHROUGH © SHARETHROUGH Analytics Initiatives
Slide 20: Our Stack
SNOWFLAKE 2017-02-03 Our Stack Amazon S3 • System of Record • mix of JSON, Parquet, and XML Snowflake Database (Launched Q4 2015) • “Dump everything here" data layer • ~100TB, or 10X the Library of Congress MicroStrategy • Defines and relates all our data • Self-serve data acquisition and exploratio