Lightning Talk: How to build a
cross-functional team to
productionise data science and
deliver automated decision-making
at pace and scale
// Harvinder Atwal
MoneySuperMarket
// Web
dunnhumby
{"previous" : "Insight Director, Tesco Clubcard"}
Lloyds Banking
Group
{"previous" : "Senior Manager, Customer Strategy and Insight"}
{"Current" : "Chief Data Officer"}
@harvindersatwal
British Airways
{"previous" : "Senior Operational Research Analyst"}
{"about" : "me"}
@gmail.com
£2.1B
SAVINGS
2018 estimate total of UK savings
1993 24.9M 13 million £356M 989
We started life
as mortgages
2000
Adults choose
to share their
data with us
Average Active
users
2018
Revenue
2018
Product
Providers
Organise around the ideal data journey not skills or technology
Data storage and Databases
Cloud file storage, NoSQL DB,Distributed file system, RDBMS, Analytical DB
Compute infrastructure and Query execution engines
VMs, Container services, and Distributed compute frameworks
Distributed SQL execution engines
Development tools, workspaces and software libraries
Data Analytics
Data exploration, Data
visualization, Data analysis,
Data science, Machine
learning, Deep learning
Reproducibility, Deployment, Orchestration and Monitoring
Output files
BI Tools
Interactive
dashboards
Web Apps
APIs
Data
Ingestion
Data
Transformation
Data
Analytics
Data
Products
Product
creation,
Customer
experience
and Business
efficiency
Use Cases
Data integration and Data processing
pipelines
ETL/ELT tools, Stream processing
MDM, Data unification, and Data preparation
Data management
Production System
Product Development System
What does the data say?
Conway's Law isn't academic
Microsoft's research found organisational structure predicted code quality better than
other measurable factors such as Code Churn, Code Complexity, Dependencies, Code
Coverage or Pre-Release Bugs
Nearly 60 percent of breakaway organizations use
cross-functional teams, versus less than a third
of the remaining respondents that do so.
#1 Team Orientation
Functional teams are organized by technical expertise, e.g.
data scientists, data engineers, BI developers, DBAs.
Optimized for maximum resource utilisation of scarce talent. No duplication of expertise across teams.
Specialised talent is allocated to projects as soon as it is available.
Database
and data
storage
team
Source
data system
owners
IT security
and data
governance
team
Data
science
team
Data
engineering
team
Data
analyst
team
Compute
infrastructure
team
IT Operations
team
Stakeholders
Use cases
Functional teams are optimised for resource utilisation not
speed
Only works well if all functional teams share goals and objectives, or they
focus on delivering self-service for other teams.
Matrix
Good Film
But terrible
organisational
structure
Domain-orientated Teams are organized by
customer, product, value-stream, segment, service,
market, etc.
Optimised for speed. A cross-functional
team is self-sufficient and has few external
dependencies.
Connecting domain-orientated teams
Spotify Model
Hub and Spoke
Chapters and Guilds.
Chapters are people with similar skills,
for example, data engineers, who meet
regularly to discuss their area of
expertise and challenges.
Guilds are communities of interest or
communities of practice who want to
share knowledge, tools, code, and
practices.
Loose Connection Tighter Connection
Central team or Centre of
Excellence ensures best
practice and consistency
Coordinator
SME, e.g. solutions
architect keeps teams
aligned on best practice
#2 Reporting Lines
Centralised reporting lines
Decentralised Integrated - Netflix model
“Highly aligned, Loosely coupled”
Works well with very large teams
Hybrid Centralized/Embedded -
AirBnB
Corporate
Centre of Excellence Business unit Function
Analytics Group Analytics Group
Analytics Project Analytics Project
#3 Skills Personas
Data
Scientist
Data
Analyst
Data
Engineer
Not all roles fit in cross-functional teams
ETL
Quality Testing
Descriptive
Analytics
Advanced
Analytics
BI
Dev
Ops
IT
Ops
DBAs
Business
Stakeholders
Operations
Sys
admin
Developers
ML
Product Managers
Core
Personas
Data Engineer Data AnalystData Scientist Team Lead
Data Platform
Administration
ML
Engineer
Supporting
Personas
Solutions Architect
DBA
Security Expert
Specialist
Tester
Technical Lead Designer
There’s a problem with cross-
functional teams
Silos within teams
#4 Cross-skill
Dash-Shaped
(Generalist)
Capable in a lot of things
but not expert in any
No I (or -) in teams
I-Shaped
(Specialist)
Expert at one thing
Poor Better Good Best
Breadth of knowledge
Depth of knowledge
T-Shaped
(Generalising Specialist)
Capable in a lot of things
and expert in one.
Pi-Shaped
(Multi-skilled)
M-Shaped
(Poly-skilled)
Analytics
Specialists
and Centre of
Excellence
Source data
system
owners
Data Management and Platform teams
(Databases, data storage, compute infrastructure, analytical tools, data governance and security,
master data management, operations, etc.)
Domain
use cases
Cross-functional domain team
(Data engineers, Data scientists, Data analysts, Stakeholder,
etc.)
Cross-functional domain team
(Data engineers, Data scientists, Data analysts, Stakeholder,
etc.)
Cross-functional domain team
(Data engineers, Data scientists, Data analysts, Stakeholder,
etc.)
Domain
use cases
Domain
use cases
Self-service access Productionise
Domain-orientated Teams (optimised for speed)
"Perfection cannot be reached because nothing's
perfect. But things can change, and so can people"
John O'Callaghan
John O'Callaghan
SHAMELESS
SELF-PROMOTION
AHEAD
// Harvinder Atwal // Web
var current: {
companyName : "MoneySuperMarket",
position : "Chief Data Officer"
};
var previous1: {
companyName : "Dunnhumby",
position : "Insight Director,"
+ " Tesco Clubcard"
};
var previous2: {
companyName : "Lloyds Banking Group",
position : "Senior Manager"
};
var previous3: {
companyName : "British Airways",
position : "Senior Operational Research Analyst"
};
{"about" : "me"}
var username = "harvindersatwal";
var linkedIn = "/in/" + username;
var twitter = "@" + username;
var email = username + "@gmail.com";

Data leaders summit 2019

  • 1.
    Lightning Talk: Howto build a cross-functional team to productionise data science and deliver automated decision-making at pace and scale
  • 2.
    // Harvinder Atwal MoneySuperMarket //Web dunnhumby {"previous" : "Insight Director, Tesco Clubcard"} Lloyds Banking Group {"previous" : "Senior Manager, Customer Strategy and Insight"} {"Current" : "Chief Data Officer"} @harvindersatwal British Airways {"previous" : "Senior Operational Research Analyst"} {"about" : "me"} @gmail.com
  • 3.
    £2.1B SAVINGS 2018 estimate totalof UK savings 1993 24.9M 13 million £356M 989 We started life as mortgages 2000 Adults choose to share their data with us Average Active users 2018 Revenue 2018 Product Providers
  • 4.
    Organise around theideal data journey not skills or technology Data storage and Databases Cloud file storage, NoSQL DB,Distributed file system, RDBMS, Analytical DB Compute infrastructure and Query execution engines VMs, Container services, and Distributed compute frameworks Distributed SQL execution engines Development tools, workspaces and software libraries Data Analytics Data exploration, Data visualization, Data analysis, Data science, Machine learning, Deep learning Reproducibility, Deployment, Orchestration and Monitoring Output files BI Tools Interactive dashboards Web Apps APIs Data Ingestion Data Transformation Data Analytics Data Products Product creation, Customer experience and Business efficiency Use Cases Data integration and Data processing pipelines ETL/ELT tools, Stream processing MDM, Data unification, and Data preparation Data management Production System Product Development System
  • 5.
    What does thedata say?
  • 6.
    Conway's Law isn'tacademic Microsoft's research found organisational structure predicted code quality better than other measurable factors such as Code Churn, Code Complexity, Dependencies, Code Coverage or Pre-Release Bugs
  • 8.
    Nearly 60 percentof breakaway organizations use cross-functional teams, versus less than a third of the remaining respondents that do so.
  • 9.
  • 10.
    Functional teams areorganized by technical expertise, e.g. data scientists, data engineers, BI developers, DBAs. Optimized for maximum resource utilisation of scarce talent. No duplication of expertise across teams. Specialised talent is allocated to projects as soon as it is available.
  • 11.
    Database and data storage team Source data system owners ITsecurity and data governance team Data science team Data engineering team Data analyst team Compute infrastructure team IT Operations team Stakeholders Use cases Functional teams are optimised for resource utilisation not speed Only works well if all functional teams share goals and objectives, or they focus on delivering self-service for other teams.
  • 12.
  • 13.
    Domain-orientated Teams areorganized by customer, product, value-stream, segment, service, market, etc. Optimised for speed. A cross-functional team is self-sufficient and has few external dependencies.
  • 14.
    Connecting domain-orientated teams SpotifyModel Hub and Spoke Chapters and Guilds. Chapters are people with similar skills, for example, data engineers, who meet regularly to discuss their area of expertise and challenges. Guilds are communities of interest or communities of practice who want to share knowledge, tools, code, and practices. Loose Connection Tighter Connection Central team or Centre of Excellence ensures best practice and consistency Coordinator SME, e.g. solutions architect keeps teams aligned on best practice
  • 15.
  • 16.
  • 17.
    Decentralised Integrated -Netflix model “Highly aligned, Loosely coupled” Works well with very large teams
  • 18.
    Hybrid Centralized/Embedded - AirBnB Corporate Centreof Excellence Business unit Function Analytics Group Analytics Group Analytics Project Analytics Project
  • 19.
  • 20.
    Data Scientist Data Analyst Data Engineer Not all rolesfit in cross-functional teams ETL Quality Testing Descriptive Analytics Advanced Analytics BI Dev Ops IT Ops DBAs Business Stakeholders Operations Sys admin Developers ML Product Managers
  • 21.
    Core Personas Data Engineer DataAnalystData Scientist Team Lead Data Platform Administration ML Engineer Supporting Personas Solutions Architect DBA Security Expert Specialist Tester Technical Lead Designer
  • 22.
    There’s a problemwith cross- functional teams
  • 23.
  • 24.
  • 25.
    Dash-Shaped (Generalist) Capable in alot of things but not expert in any No I (or -) in teams I-Shaped (Specialist) Expert at one thing Poor Better Good Best Breadth of knowledge Depth of knowledge T-Shaped (Generalising Specialist) Capable in a lot of things and expert in one. Pi-Shaped (Multi-skilled) M-Shaped (Poly-skilled)
  • 26.
    Analytics Specialists and Centre of Excellence Sourcedata system owners Data Management and Platform teams (Databases, data storage, compute infrastructure, analytical tools, data governance and security, master data management, operations, etc.) Domain use cases Cross-functional domain team (Data engineers, Data scientists, Data analysts, Stakeholder, etc.) Cross-functional domain team (Data engineers, Data scientists, Data analysts, Stakeholder, etc.) Cross-functional domain team (Data engineers, Data scientists, Data analysts, Stakeholder, etc.) Domain use cases Domain use cases Self-service access Productionise Domain-orientated Teams (optimised for speed)
  • 27.
    "Perfection cannot bereached because nothing's perfect. But things can change, and so can people" John O'Callaghan John O'Callaghan
  • 28.
  • 31.
    // Harvinder Atwal// Web var current: { companyName : "MoneySuperMarket", position : "Chief Data Officer" }; var previous1: { companyName : "Dunnhumby", position : "Insight Director," + " Tesco Clubcard" }; var previous2: { companyName : "Lloyds Banking Group", position : "Senior Manager" }; var previous3: { companyName : "British Airways", position : "Senior Operational Research Analyst" }; {"about" : "me"} var username = "harvindersatwal"; var linkedIn = "/in/" + username; var twitter = "@" + username; var email = username + "@gmail.com";