Data Teams: A Unified Management Model for Successful Data-Focused Teams

Table of Contents
About the Author
About the Technical Reviewer
Acknowledgments
Introduction
Part I: Introducing Data Teams
	Chapter 1: Data Teams
		Big Data and Data Products
			The Terrible 3s, 4s, 5s…​
			The Can’t Definition
			Why Management Needs to Know the Definition of Big Data
		Why Is Big Data So Complicated?
			Data Pipelines and Data Products
		Common Misconceptions
			“It’s All Just Data”
			“Isn’t This Just Something Slightly Different from…​?”
				Business Intelligence
				Data Warehousing
				Operations
				Software Engineering
		Why Are Data Teams Needed for Big Data?
		Why Some Teams Fail and Some Succeed
		The Three Teams
			Data Science
			Data Engineering
			Operations
		Why Are Three Teams Needed?
			Three Teams for Small Organizations
		What Happens If an Organization Doesn’t Manage Correctly?
	Chapter 2: The Good, the Bad, and the Ugly Data Teams
		Successful Data Teams
			What Big Data Success Looks Like
			What Big Data Failure Looks Like
			What Underperforming Projects Look Like
			What Happens When a Team Is Missing
			Figuring Out Value Generated
			Issues with Scale in Data Science
			Automate As Much As Possible
Part II: Building Your Data Team
	Chapter 3: The Data Science Team
		What Skills Are Needed?
			Math
			Programming
			Distributed Systems
			Communication
			Domain Knowledge
				What characterizes the problem?
				How appropriate is the data?
				What other data do we need?
		Technical Debt in Data Science Teams
		Hiring and Training Data Scientists
			The Barriers to Retraining
			Upgrading the Skills of Data Scientists
			Finding Data Scientists
		Meeting the Needs of Data Scientists
			Introducing Software Engineering Practices
			Too Much Process Stifles Progress
	Chapter 4: The Data Engineering Team
		What Skills Are Needed?
			Distributed Systems
			Programming
			Analysis
			Visual Communication
			Verbal Communication
			SQL
			Schema
			Domain Knowledge
			Other Important Skills
				Data Governance
				Data Lineage
				Data Metadata
				Discovery Systems
		Levels of Expertise
			New Data Engineers
			Qualified Data Engineer
			Veteran
		Further Specialization
		Should the Data Engineering Team Focus Only on Big Data?
		Common Misconception
			Why Data Engineering Isn’t Just Data Warehousing
			Why a Data Engineer Isn’t a Data Scientist
			Why Data Engineering Is More Than Data Wrangling
		The Relationship Between a Data Engineering Team and an Existing Data Science Team
		Retraining Existing Staff
			Software Engineers
			SQL-Focused Positions
		The Role of Architects
			Placement in the Organization
	Chapter 5: The Operations Team
		The Special Challenge of Operations on Distributed Systems
		Job Titles for Operations Teams
		What Skills Are Needed?
			Hardware
			Software/Operating Systems
			Distributed Systems
			Troubleshooting
			Security
			Data Structures and Formats
			Scripting/Programming
			Operationalization Best Practices
			Monitoring and Instrumenting
			Disaster Recovery
		Establishing Service-Level Agreements
			Batch SLAs
			Real-Time SLAs
			Specific Service/Technology SLAs
			Organization Code and Deployment SLAs
		Typical Problems Faced
			Your Organization’s Code
			Data and Data Quality
			Framework Software
			Hardware
		Staffing the Operations Team
			The Need for Specialized Training for Big Data
			Retraining Existing Staff
				Operations
				DevOps Staff
			Why a Data Engineer Isn’t a Good Operations Engineer
		Cloud vs. On-Premises Systems
			Managed Cloud Services and Operations
	Chapter 6: Specialized Staff
		DataOps
			The Trade-Offs Made in DataOps
			Finding Staff for DataOps
			The Value of DataOps
			Relationship Between the DataOps and Data Engineering Teams
			When to Start Doing DataOps
		Machine Learning Engineers
			Finding Machine Learning Engineers
			Where to Locate Machine Learning Engineers
Part III: Working Together and Managing Data Teams
	Chapter 7: Working as a Data Team
		Getting People to Work Together
			Staffing Ratios
			Should Data Teams Be Separate or Together?
			High-Bandwidth Connections
			An Iterative Process
		Politics, Expectations, and Egos
			Data Project Velocity
			Create a Crawl, Walk, Run Plan
			Spreading the Love
			Communicating the Value of Data Engineering, Data Science, and Operations
			Data Scientists Need the Contributions of Other Teams
			Data Hoarding
		Death by a Thousand Cuts
			The Curse of Complexity
			Going Off the Beaten Path
		Technical Debt
	Chapter 8: How the Business Interacts with Data Teams
		How Change Can Be Accomplished
			Pushing from the Top Down
			Middle Pressure Up and Down
			Bottom Up
		How Should the Business Interact with Data Teams?
			Case Study: Medical Insurance Domain Knowledge
				Medical
				Insurance
				Billing Systems
				Data Systems
			Switching from Software as a Product to Data as a Product
			Symptoms of Insufficient or Ineffective Interaction
			Working with the QA Team
			Working with Project Managers
		Funding and Resources
			Staffing
			Software and Hardware
			Cloud
		Topics for Interaction
			Data Strategies
			Risks and Rewards
			Managing and Creating Realistic Goals
			Use Cases and Technology Choices
			Speaking Truth to Power
			Executive-Level Attention
		Dealing with Uncertainty
			Data Sources Are Changing
			Output of Models Is a Level of Certainty
		Don’t Fear the Reaper’s Data Products
		Don’t Forget the People Side
			Data Warehousing/DBA Staff
			SQL Developers/ETL Developers
			Operations
			Business Intelligence and Data Analyst Teams
		Establishing Key Performance Indicators (KPIs)
			Data Science Team
			Data Engineering Team
			Operations
	Chapter 9: Managing Big Data Projects
		Planning the Creation and Use of Data Products
			One-Off and Ad Hoc Insights
			Exposing Data Products
		Assigning Tasks to Teams
			Data Engineering, Data Science, or Operations?
			Collaboration
			Rectifying Problems with Existing Teams
			Effects of Using the Wrong Team for a Task
		Long-Term Project Management
		Technology Choices
			A Mental Framework for Adding or Choosing Technologies
			Reining in Technology Choices
			Programming Language Support
				Languages Used by Data Scientists
				Languages Used by Data Engineers
				Exposing Data Across Languages
			Project Management
		When It All Goes Wrong
		Look Out for Nth Order Consequences
	Chapter 10: Starting a Team
		Establishing New Teams
			The First Data Engineer
			The First Data Scientist
			Operations
			The Critical First Hire
		Location and Status of the Data Team
			Embedding the Data Team in Business Unit
			Hub and Spoke
			Center of Excellence
			Data Democratization
		Getting the Team Right from the Start
			Data Science Startups
			Companies Lacking a Data Focus
			Small Organizations and Startups
				Consequences of a Bad Hire
				Consequences of Hiring Only Junior People
			Medium-Sized Organizations
			Large Organizations
		What Should the Reporting Structure Look Like?
			VP of Engineering
				CTO
				CEO
				CIO
				VP of Marketing
				VP of Product
				VP of Finance
				Chief Data Officer (CDO)
				VP of Data
		Working with Human Resources Departments
			This Isn’t Just a Focus on Titles
			Adding New Titles
				Title of Data Scientist
				Title of Data Engineer
				Title of Operations Engineer
			Focus on Technology Skills
	Chapter 11: The Steps for Successful Big Data Projects
		Presteps
			Look at Data and Data Problems
			Identify Business Needs
			Is the Problem a Big Data Problem?
			Decide on an Execution Strategy
		Between Planning and Starting
			Find Some Low-Hanging Fruit
			Waiting for Data
			Deeply Understand the Use Cases
		Starting the Project
			Getting Outside Help
				Training
				Consulting
			Choose the Right Technologies
			Write the Code
			Create the Compute Cluster
			Create Metrics for Success
		Iterative Steps
			Evaluate
			When to Repeat
	Chapter 12: Organizational Changes
		Working with the Old Guard
		What to Do with Previous Teams?
		Innovation and Progression in Data Engineering and Data Science
		Assessing Skills Gaps
			Assessing Levels of Expertise
			How to Perform a Gap Analysis
			Interpreting Your Gap Analysis
			Ability Gaps
		Hardware Changes
			Cloud Usage
			Purchasing Clusters
			Systems for Big Data Teams
				Development Hardware
				Test and QA Hardware
			Getting Data to Data Scientists
	Chapter 13: Diagnosing and Fixing Problems
		Stuck Teams
			The team says they’ll get things done, but they said the same thing a month ago
			Whenever I give the team a task they come back and say it isn’t possible
			I can’t tell the difference between the team making progress and being stuck
			We have trouble with small data systems, and we’re having even more trouble with big data
		Underperforming Teams
			Whenever I give the data team a task, the team creates something that doesn’t work
			Whenever I give the team a task, they create something that isn’t really what I asked for
			The team can do elementary things, but they can’t ever do something more complicated
		Skills and Ability Gaps
			How can I tell the difference between a skill gap and ability gap?
			Is the team having trouble learning how to program?
			The team is coming from a small data background and having difficulty learning big data and distributed systems
			My team says that you’re wrong and that the task is easy
			Why can’t I just hire some beginners and have them create this project?
			We built out a distributed systems cluster, and no one is using it
		Out-of-Whack Ratios
			My data scientists keep complaining that they have to do everything themselves
			My data scientists keep dropping or stopping projects
			The data engineers’ analysis isn’t very good
		Project Failures and Silver Bullets
			We’ve tried several projects, and none of them went anywhere
			We’ve given up on big data because we can’t get any ROI
			We’ve tried cloud and failed; now we’re failing with big data
			This is really hard; is there an easier way to get it done?
			Is there any simple way or shortcuts to these problems?
			We hired a consultancy to help us, but they aren’t getting it done
			We followed everything our vendor told us to do, and we’re still not successful
			We defined our data strategy, but nothing is being created
			Our data science models keep failing in production
		The Holy Grail
			We copied someone’s architecture, and we’re not getting the same value
			We have a really ambitious plan, and we’re having trouble accomplishing it
		The Software or Data Pipeline Keeps Failing in Production
			We keep on having production outages
			The data keeps on bringing our system down, and we can’t stop it
			It takes way too long to find and fix issues in code and production
Part IV: Case Studies and Interviews
	Chapter 14: Interview with Eric Colson and Brad Klingenberg
		About This Interview
			Background
			Starting Point
			Growth and Hiring
			The Primary Division into Data Science and Platform Teams
				Data Science Team
					Merchandising Algorithms
					Styling Algorithms
					Client Algorithms
					Operations Algorithms
					Customer Service Algorithms
				Platform Team
			Bottom-Up Approach
			Project Management
			The Competitive Edge of Data
			Advice to Other Companies
			Lessons Learned in Hindsight
	Chapter 15: Interview with Dmitriy Ryaboy
		About This Interview
			Background
			Hiring at Twitter
			Challenges of Data and Analytics
			Task Ownership Structure
			The Difficulty of Choosing Technologies and Creating Distributed Systems
			Data Engineers, Data Scientists, and Operations Engineers
			Project Management Framework
			Business and Data Team Interactions
			Keys to Success with Data Teams
	Chapter 16: Interview with Bas Geerdink
		About This Interview
			Background
			ING’s Data Teams
			ING Organizational Structure
			Project Management Frameworks
			Using Data Science in Banking
			KPIs for Data Teams
			Advice for Others
	Chapter 17: Interview with Harvinder Atwal
		About This Interview
			Background
			Team Structure
			Removing Barriers and Friction
			Project Management Frameworks
			Team KPIs and Objectives
			Changes in Data Teams
			Advice to Others
	Chapter 18: Interview with a Large British Telecommunications Company
		About This Interview
			Background
			Starting the Initiative
			Working with the Business
			Data Scientists and Data Engineers
			Supporting Models for the Enterprise
			Moving from Proof of Concept to Production
			Creating Enterprise Infrastructure and Operations
			Project Management Frameworks
			Advice for Others
	Chapter 19: Interview with  Mikio Braun
		About This Interview
			Background
			Organizational Structure
			Socializing the Value of Doing Machine Learning
			Definitions of Job Titles
			Reducing Friction
			Project Management Frameworks
			KPIs
			Improving Data Scientists’ Engineering Skills
			Integrating Business and Data Teams
			The Differences Between European and US Companies
			Advice to Others
Index