Snowflake Inc. · NYSE:SNOW

Snowflake — Investor Presentation (November 2019)

13 slides from Snowflake's investor presentation, November 2019

  1. Slide 1: Presented by Sandro Frattura, B.I. Data Architect

    Presented by: Sandro Frattura B.I. Data Architect © 2018 Snowflake Computing Inc. All Rights Reserved. 1

  2. Slide 2: About Hubspot

    About Hubspot Who is Hubspot? Hubspot helps small-to-medium sized business GROW BETTER ○ inbound marketing software and automation ○ sales management software (CRM) ○ website content management ○ SEO optimization ○ Blogging & social media integration Our mission to help business attract leads and tu

  3. Slide 3: About Hubspot: We are not a gym!

    About Hubspot WE ARE NOT A GYM! © 2018 Snowflake Computing Inc. All Rights Reserved. 3

  4. Slide 4: Life Before Snowflake

    Life Before Snowflake Scenario Bring together disparate data sources into one unified data warehouse Everything that we did left our jaw on the WAIT Pain Points YOUR table. ‘Wait – we’ve never done anything like ● Resource contention that.’ Or, ‘How did that just run so fast.’ We Club Cartesian TURN!

  5. Slide 5: Data Flow From 30,000 Feet

    Data Flow From 30,000 Feet Selectively, data gets “pushed back” to the source systems Dozens of proprietary MySQL data sources SalesForce Snowflake Data Science Netsuite S3 Jira Amplitude Workday Looker for data exploration & “canned” reports Etc., HUBSPOT API © 2018 Snowflake Computing Inc. All Right

  6. Slide 6: What's in the Box???

    WHAT’S IN THE BOX??? DATA LAKE (“EL” of ETL) DATA WAREHOUSE (HubsHouse™) (“T” of ETL) ● 90+ schemas ● highly curated data ● raw data ○ pre-joined ● one per source DB ○ transformed ○ aggregates ○ filtered (I.E. “latest” or “relevant”) ○ column-ized from lake JSON ○ slowly changing dimensions Unified da

  7. Slide 7: Isolatable/Scalable Computing

    Isolatable/Scalable Computing Before After fighting for single resource ETL dedicated resources, right-sized LOOKER DATA USERS SCIENCE ETL LOOKER DATA USERS SCIENCE Assign computing Eliminate power purposefully bottlenecks Measure usage and Scale dynamically © 2018 Snowflake Computing Inc. All Rights

  8. Slide 8: Miscellaneous Goodness

    Miscellaneous Goodness ** Query Profiler ❖ Natively support of JSON (et al): You can query a single-level JSON object as efficiently as transforming it into rows/columns. For more complex JSON (nested, arrays, lists, time-series), you should FLATTEN the data in your transformation layer ❖ Merge: The m

  9. Slide 9: Tip #1: Migrate Immediately

    Tip #1: Migrate Immediately ❖ DATA ENGINEERS: Changing all of your pipelines to send data to Snowflake will take some engineering effort ➢ you might want to address some tech debt while you rewrite your ETL code ❖ DATA ANALYSTS: some functions from legacy DB need to be rewritten. This takes time ➢ yo

  10. Slide 10: Tip #2: Cluster Large Data Sets

    Tip #2: Cluster Large Data Sets Regardless of #nodes, CPU speed, amount of RAM etc., the most important aspect of query performance on large data sets is data organization. In Snowflake, this means applying cluster keys on your large tables and then maintaining the organization by reclusting regularl

  11. Slide 11: Tip #3: Manage Warehouse Sizes

    Tip #3: Manage Warehouse Sizes ● SHUT IT DOWN: make sure warehouses auto-suspend after periods of inactivity ● SUPERSIZE ME: programatically upsize (on-the-fly) your warehouse for large ETL operations ● NO “BORED” WAREHOUSES: check & monitor usage. If you have very few queries, scattered throughout a

  12. Slide 12: Closing Thoughts

    Closing Thoughts How we brought our data nation together with Snowflake • Have all of our data (raw and curated) in one place • Give everyone access to data, eliminating resource conflict • Leverage the cloud © 2018 Snowflake Computing Inc. All Rights Reserved. 12

  13. Slide 13: Thank you!

    Thank you! © 2018 Snowflake Computing Inc. All Rights Reserved. 13

Snowflake — Investor Presentation (November 2019) on 6ix