byo/diy analytics platform (measurecamp presentation by clancy childs)
DESCRIPTION
The slides accompanying Clancy Childs' talk at Measurecamp V (2014) in London. Might be missing a lot if you weren't at the session, but basically covering some of the design decisions, pitfalls, technology choices and requirements when choosing to build your own analytics / eventing platform and data warehouse.TRANSCRIPT
BYO/DIY Analy,cs Pla1orm
Clancy Childs Founding Partner, Inderus (Google Analy,cs Refugee)
[email protected] @cchilds -‐ @inderus
My Past Life
WHY would you ever WANT to build your own Analy,cs Solu,on?
“See all that stuff in there, Homer? That’s why your robot never worked.”
Tools Are Generally Good Things
The Right Tool for the Job
…but eagerness to use a tool can be at the cost of geWng the job done correctly
Why Some Companies “Roll Their Own” Analy,cs
• End-‐to-‐end customisa,on (collec,on, processing and repor,ng)
• Complete control and governance of data
• Integra,ons with other opera,onal and repor,ng systems
• Ability to make their own mistakes on their own terms
An Insanely Simplified Analy,cs Processing Model
Collection
Processing
Enrichment
Extraction VisualisationStorage
An Insanely Simplified Analy,cs Processing Model
Collection
Processing
Enrichment
Extraction VisualisationStorage
Collec,on
• First vs. Third Party beacons
• Client vs. Server Side collec,on (over GTM’d?)
• Opera,onal By-‐Products (Logging)
• Snowplow, Logstash, Kinesis, Kaaa(?)
An Insanely Simplified Analy,cs Processing Model
Collection
Processing
Enrichment
Extraction VisualisationStorage
Processing and Enrichment• Sessioniza,on and User S,tching
• Processing Schedules and Goals(Lambda Architecture?)
• Reprocessing vs “Golden” Immutability
• Captured Foreign Keys (gclid, IP address, etc.)
• MapReduce, Storm, Flume…
An Insanely Simplified Analy,cs Processing Model
Collection
Processing
Enrichment
Extraction VisualisationStorage
Storage
An Insanely Simplified Analy,cs Processing Model
Collection
Processing
Enrichment
Extraction VisualisationStorage
Query/Extract and Visualize
• SQL is generally much easier than anything else.
• Prototyping with Pandas / R
• Produc,on Dashboarding with Visualiza,on tools
• Char,o, Looker, Tableau, Klipfolio, Legronic, others?