1. The New Gig Economy:
From Best Practices to Best Friends
June 2017
Benjamin Reid
CEO
@ElasticitiInc
2. The Gig Economy is Here
While AirBnb and Uber are the poster-children the
concept remains the same:
Marketplace of resources and opportunities
On-demand, variable duration
4. Usually Areas of High Risk
Aspirational project to advance company / career
Defensive project to keep up with competition
Critical area to maintain accuracy and velocity
Get the most out of an investment (people / tech)
5. Five Disciplines for Success
Find the Fit
Commit to the
Engagement Mode
Stay Goal Oriented
Think Extensibility
Formalize
Communication
7. Find the Fit - Resource
Demonstrated, relevant experience
Project management style experience
Communication style and expectations
8. Find the Fit - Team
Think Hats vs bodies
Client Consultant
Data Engineering Roger Jing
Business Lead Sarah Rachel
Data Viz Jing
Math/Stats Sarah Jing
Project Sunil Harry
Account/Sponsor Denise Harry
9. Why? The right profile leads to:
Quality of solution
Quality of dialogue
Velocity
Maintainability
Worth getting right!
11. Agile vs Waterfall
Set goals and work backward
Specific deliverable vs. thematic and iterative
Commit to the mode and the Project Management
approach
Waterfall = BRUF
Research, Plan, Ratify
Agile = North Star
Set shorter-term goals
Seek elements that will provide valuable feedback
14. Set Quantitative Goals
If iterating an existing model, how much better should it be?
How much time is allotted to achieve the goal?
If creating new models, what should threshold performance
be?
Data-integration projects should have their own tie-out
metrics
Data changes, but you can document why
16. Keep it Tidy
Create modular code. Future generations will thank you
Embrace functional coding best practices
Document as you go, not at the end
Refactoring is acceptable!
Perform real-world simulations frequently
Isolate and control for changes (environment, format,
parameters etc.)
17. Data Science Collaboration
Best Practices
Create and share packages
Use RMD to explain to end users
Push functions to a separate file
One task per script
Isolate parameters in YAML files
20. Prepare for Post-Launch Support
Are the resources in place to support the
initiative post-launch?
Do they have the skills or can they acquire them?
How much knowledge transfer is needed (run vs iterate)?
Do they have the time to pick it up?
How much documentation is expected? And who will leverage it?
21. Why are we doing this again?
Connect with sponsoring
executive at least quarterly
Ensure their success criteria is
revalidated
Deliver critical, candid updates
22. Case Study 1 – Large Broadcaster A
Six Months to Initial Launch
Team of six (3 client / 3 contract – all FT)
Address gaps in content distribution data, normalize data
Team – data engineers, data viz, PM, exec sponsor
Mode – Agile, sort of, very collaborative
Goals – Data accuracy was primary, velocity secondary
Extensible – sacrificed for speed
Comm – Daily/weekly/monthly. Very formal
23. Case Study 2 – Large Broadcaster B
Four Months to Initial Launch
Team of four (2 client / 2 contract – all PT)
Create Big Data environment for improved analytics
Team – data engineer, PM, exec sponsor
Mode – Agile, sort of, moving to Waterfall
Goals – Velocity was primary, data accuracy secondary
Extensible – highly modular code and architecture
Comm – Weekly/quarterly. Semi formal
24. Client A Client B
Contrasting Project Styles and Experiences