Deploying a Data Sciences Team -- The Promise and the Pitfalls

2,380 views
2,200 views

Published on

Intuit formed its first Data Sciences group in 2008. From inception until 2012, Intuit’s data scientists operated from a central organization and leveraged their unique expertise like “consultants” assigned to a specific project and “client” from the business unit. In October, 2012, Intuit began piloting an “embedding” strategy that places a data scientist deep within a business unit product team to use their expertise as part of the team rather than just “consulting” on a project. Since then, the embedding program has grown and currently 6 data scientists/analysts from the central organization have joined product teams across the company. Success has been mixed and in this talk we will describe the pros and cons of the two ways of deploying data scientists within the organization, as well as the key ingredients we found are necessary for great outcomes.

Published in: Technology
0 Comments
2 Likes
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
2,380
On SlideShare
0
From Embeds
0
Number of Embeds
31
Actions
Shares
0
Downloads
18
Comments
0
Likes
2
Embeds 0
No embeds

No notes for slide

Deploying a Data Sciences Team -- The Promise and the Pitfalls

  1. 1. Simplify the business of life Deploying a Data Sciences Team: The Promise and the Pitfalls Diane Chang, PhD Senior Data Scientist
  2. 2. … or Why I love embedding 2
  3. 3. 3
  4. 4. You’ve hired them, now how do you deploy them? 4
  5. 5. • Centralized team • Resident in the business • Embedded in business 5
  6. 6. Data Science 6
  7. 7. My Intuit experience • 5 years as a data scientist • First 4 years in “centralized team” mode • Last year I was embedded for 10 months … and I loved it! 7
  8. 8. My concerns with centralization • • • • 8 Single point of contact Few direct interactions Single source for context Feel less a part of the team
  9. 9. The first embedding experiment • Big Data for the Little Guy • First volunteer 4 +4 +4 +4 +4 +4 +4 +4 +4 +4 3 months 3 months 4 months The result: A new business with data in its DNA! 9
  10. 10. Being one of the team • Direct communications – No “geek speak” • Hallway conversations • Flash mob • Extra context -> better product Plus: • Maintained connection with data science team 10
  11. 11. Why I love embedding • • • • 11 New data “believers” Significant impact on a new product Learned a lot Made new friends
  12. 12. And what I learned… • It’s easy to get disconnected • Enter with an exit strategy • Demand can be overwhelming • Two homes can feel like no home 12
  13. 13. Hey product team - are you ready? • Well-defined problem with significant business value • Data is available • Committed bandwidth from team • Welcoming – have a buddy 13
  14. 14. Data scientist - do you have what it takes? • • • • • Open to change Value learning Easily develop personal relationships Adaptable Translate technical to business terms ∑  ≈ ∏ 14
  15. 15. Since the first experiment… • We are growing the program – A dozen embedding rotations • We are still experimenting – Pods: a team not an individual • Can play a larger strategic role • Can offer a variety of skills 15
  16. 16. And me? • Doing a rotation back in the center • Improving tools and products that benefit many product teams • Increasing my skills and knowledge 16

×