Data @Rent the Runway - Jay Grossman 2017-06-06 - Snowflake

Page created by Norman Fisher
 
CONTINUE READING
Data @Rent the Runway - Jay Grossman 2017-06-06 - Snowflake
Data @Rent the Runway

      Jay Grossman
       2017-06-06
Data @Rent the Runway - Jay Grossman 2017-06-06 - Snowflake
Women’s fashion for events
     is expensive!
Data @Rent the Runway - Jay Grossman 2017-06-06 - Snowflake
Items get worn once or few times.
Data @Rent the Runway - Jay Grossman 2017-06-06 - Snowflake
Time for a cost effective solution:
Data @Rent the Runway - Jay Grossman 2017-06-06 - Snowflake
With thousands of styles to choose
Data @Rent the Runway - Jay Grossman 2017-06-06 - Snowflake
Delivered to your home
Data @Rent the Runway - Jay Grossman 2017-06-06 - Snowflake
Rent the Runway delivers:

            • Rent designer fashion for
              4 or 8 days

            • Subscribe monthly to fashion
              with 3 garments or accessories
              (similar to Netflix model)
Data @Rent the Runway - Jay Grossman 2017-06-06 - Snowflake
Data drives our company!
Data @Rent the Runway - Jay Grossman 2017-06-06 - Snowflake
RentTheRunway.com

•   Desktop + Mobile Web + iOS platforms
•   Homegrown clickstream and experiment frameworks
•   Demand Analysis and Pricing Optimization
•   Personalization and Recommendations
Data @Rent the Runway - Jay Grossman 2017-06-06 - Snowflake
Reservation allocation similar to
      hotels and airlines
Retail showrooms in 6 cities

•   Homegrown POS system for rentals
•   Optimize product merchandising for showroom
•   Staffing optimizations and sales reporting
•   Growth / expansion analysis
100+ Customer Service reps
24/7 warehouse ops with custom
WMS to manage reverse logistics
Dry cleaning at scale
Planning, Buying, Merchandising

•   More complex than retail planning/buying cycles
•   We carry 400+ Designers
•   Demand modeling for Event and Subscription models
•   Seasonal buys and replenishment needs
Marketing & Acquisition
Rent the Runway Data Stack
Why Snowflake?
What sounds better?

• Manage a Hadoop cluster

• Manage a Vertica cluster

• Use Data Warehouse as a Service

• We prefer to focus resources on delivering business value over system
  maintenance.
Why Snowflake?

• Even with training, managing an on-premise Vertica implementation (black box)
  has been challenging.
Why Snowflake?
Variant data type makes NoSQL less painful.
DW          Data type        Query

Vertica     varchar(64000)   Select * from table where
                             trim(regexp_substr(json_data,
                             'referrer":"(.*?)"',1,1,'b',1))),'‘) = ‘blah’

Snowflake   VARIANT          Select * from table where
                             json_data:referrer = ‘blah’
Why Snowflake?
Our analysts know SQL + Analytic functions.
Snowflake learning curve is minimal.

• Priority is support for 14 person analytics
  team are the primary consumers (+ marketing,
  merchandising, finance, engineering teams).
• Fairly clean integration with 100’s of Tableau
  reports.
• Kafka connector support coming in 2018.
Why Snowflake?
ERROR 3007: DDL statement interfered with
this statement

• This happens regularly when Vertica tries to
  sync data between nodes.
  – It SUCKS when it happens in the middle of large
    ETL jobs at 3:00 AM.
• Process isolation in snowflake means we don’t
  have to worry about this horrible error!
Why Snowflake?
Snowflake looks more like a data lake!

• With Vertica, our license allows a capped
  amount of data in production.
  – We either need to pay more $$$ or remove data
    to explore new data sets.
• Since Snowflake’s is architected on top of S3,
  we can scale data capacity up and down as
  needed.
Why Snowflake?
Scale up beefy processes independently.

• We had challenges executing complex daily
  ETL workflows within a 24 hour period.
• Snowflake’s data warehouses allow us to allot
  compute for intensive processes independent
  of other activities.
Snowflake Wrap Up
1.   Cheaper to scale
2.   Easier to scale
3.   We don’t have to be DBA’s
4.   Small learning curve
Questions???
You can also read