Everyday Data Science

Paul Laughlin
Chief Blogger, CustomerInsightLeader.com
Founder, LaughlinConsultancy.com
“Everyday Data Science”
Masterclass

My background in creating data,
analytics & Data Science teams
2
❖ Created and lead data, analytics & data
science teams for all the major insurance
brands, products & channels used by
Lloyds Banking Group over 13 years
❖ Added over £11m incremental profit to
bottom line annually
❖ Pioneered work with FCA on Behavioural
Economics in comms
❖ Developed capability in team of 44 &
mentored next generation of leaders

The way spend my time now
reﬂects needs of Data Science
3
“Helping businesses make money from customer insight”
A selection of the clients I work with to embed these skills…

The 2nd biggest predictor of
whether you get value from this…
4
Goals:
What do you want to get
out of this morning?

01 All around us, but let’s get clearer
What is it?
02 Some exciting & concerning applications
How is it being used?
04 How could you learn the coding needed?
Coding Opportunities
03 How can you spot new data & uses?
Data Opportunities
05 What is needed to get started?
Getting ready
5
09:45
10:30
11:30
12:00
12:30

Data Science - What is it?
(Part 1 of 5)
6

Rarely a week goes past without
Data Science being in the news
7

Some applications are showing
great potential to beneﬁt society
8

Other applications raise ethical
concerns so we should all be aware
9

Pause to Think:
How has Data Science
already impacted you?

But what is Data Science & where
has it come from?
11Source: CapGemini.com

Probably the most popular
deﬁnition of Data Science
12
Source: DrewConway.com Source: oralytics.com Source: datacommunitydc.org

Breadth of Skills needed is
recognised for Data Scientists
Drew Conway “The Data Science Venn Diagram” March 2013
<http://drewconway.com/ zia/2013/3/26/the-data-science-venn-diagram>
Page 11 of 59
Figure 3. ICTprocessstagesalignedwiththe organisationalproductionworkflow (as usedine-CF3.0)
Figure 4 illustrates the multi-purpose use of the European e-Competence Framework within ICT organisations.
The e-CF has a multidimensional structure and is flexible in using for different purposes, it can be easy adopted
for organisation specific model and roles. The e-CF3.0 is used for job-profiles definition in CWA 16458 (see [9]
and EDSF DSPP document [4]) that are linked to the organisational processes what creates limitations for cross-
organisational professional profiles and roles such as Data Scientist. However, combining competences from
different competence areas and using them as building blocks can allow flexible job-profiles definition. This
enables the derived job-profiles to be easily updated by changing set of competences related to profiles without
the need to restructure the entire profile.
Figure 4. e-CF3.0 structure and use for definition of the job profile definition and training needs.
EU EDISON Data Science Framework (2017)
13

but too much of focus is on
developing Technical Skills
EDSF Release 2: Part 1. Data Science Competence Framework (CF-DS)
Table 4.2. Identified Data Science skills related to the main Data Science competence groups
SDSDA
Data Science
Analytics
SDSENG
Data Science
Engineering
SDSDM
Data Management
SDSRM
Research Methods
and Project
Management
SDSBA
Business Analytics
SDSDA01
Use Machine Learning
technology,
algorithms, tools
(including supervised,
unsupervised, or
reinforced learning)
SDSENG01
Use systems and
software engineering
principles to
organisations
information system
design and development,
including requirements
design
SDSDM01
Specify, develop and
implement enterprise
data management and
data governance
strategy and
architecture, including
Data Management Plan
(DMP)
SDSRM01
Use research methods
principles in developing
data driven applications
and implementing the
whole cycle of data
handling
SDSBA01
and Business
Intelligence (BI)
methods for data
analysis; apply
cognitive
technologies and
relevant services
SDSDA02
Use Data Mining
techniques
SDSENG02
Use Cloud Computing
technologies and cloud
powered services design
for data infrastructure
and data handling
services
SDSDM02
Data storage systems,
data archive services,
digital libraries, and their
operational models
SDSRM02
Design experiment,
develop and implement
data collection process
SDSBA02
Apply Business
Processes
Management (BPM),
general business
processes and
operations for
EDSF Release 2: Part 1. Data Science Competence Framework (CF-DS)
Table 4.3. Required skills related to analytics languages, tools, platforms and Big Data infrastructure 6
DSDALANG
Data Analytics
and Statistical
languages and
tools
DSADB
Databases and
query
languages
DSVIZ
Data/Applicatio
ns visualization
DSADM
Data
Management
and Curation
platform
DSBDA
Big Data
Analytics
platforms
DSDEV
Development and
project
management
frameworks,
platforms and tool
DSDALANG01
R and data analytics
libraries (cran,
ggplot2, dplyr,
reshap2, etc.)
DSADB01
SQL and
relational
databases (open
source:
PostgreSQL,
mySQL, Nettezza,
etc.)
DSVIZ01
Data visualization
Libraries
(mathpoltlib,
seaborn, D3.js,
FusionCharts,
Chart.js, other)
DSADM01
Data modelling
and related
technologies (ETL,
OLAP, OLTP, etc.)
DSBDA01
Big Data and
distributed
computing tools
(Spark,
MapReduce,
Hadoop, Mahout,
Lucene, NLTK,
Pregel, etc.)
DSDEV01
Frameworks: Python,
Java or C/C++, AJAX
(Asynchronous
Javascript and XML),
D3.js (Data-Driven
Documents), jQuery,
others
Source: EU EDISON Data Science Framework (2017)
14

Leaders consistently identify
need for People Skills
15

A 9-step model for effective use
of Softer Skills in Data Science
Question
Data Analysis Insight
Planning & Design
Presentation &
Distribution
Solution
Buy - in Sign - off
“Contracting” translating
business questions into
actionable, analytical
terms
“Delivering”
expressing analysis &
insight in actionable
business terms
Addressing business
need
Transparency of
activity
Engagement with key
stakeholders
16

Don’t forget the Science word,
it means working in new ways
17

All this means that Data Scientists
need freedom to work differently
18
https://docs.microsoft.com/en-us/azure/machine-learning/team-data-science-process/lifecycle

If organisation is conducive this
can enable new ways of working
19
Day One: Empathise
Day Two: Deﬁne
Day Three: Ideate
Day Four: Prototype
Day Five: Test

Pause to Think:
Is your organisation
ready to hire a Data
Scientist?

Data Science applications
(Part 2 of 5)
21

Data Science is being applied
to a wide variety of problems…
22
C2 General
Whether you’re a Retailer, Real Estate, CPG
or Tourism organization, Vodafone Analytics
allows you to use anonymized mobile data
insights to drive decisions on:
● Site Planning
● Geomarketing
● Performance Management
● Investment Analysis
● Smart tourism
New digital
data products
Vodafone Analytics – Location Intelligence

23
What is attribution modelling?
Each marketing channel
has an eﬀect on the
customer.

Attribution modelling is
to determine these
eﬀects.
https://github.com/5ri/WiML2017/blob/master/Skyscanner-MultiTouchAttribution-Poster.pdf

24
12
Proactively monitor key performance metrics across all timeseries and dimensions,
autonomously automating the generation of insights
LIBERTY GLOBAL – PERFORMANCE MANAGEMENT
DATA LEADERS SUMMIT | October 18, 2018 | appli Performance Management Application UI Design
Retail Store Gross Additions
Flanders Retail Locations
West Flanders Weekly Retail Gross Additions
Retail Store Weekly Sales per Store
Retail Net New ARPU
109 119 120
Retail gross additions are declining…
16.7
Sample Data
‒ Disguised ‒
Copyright 2018 appli. Inc. Patents pending (USPTO #62562910, #62625645).
All rights reserved.
SAMPLE USE CASES
PERFORMANCE MANAGEMENTPERFORMANCE MANAGEMENT MARKETING ANALYTICSMARKETING ANALYTICS PREDICTIVE MODELINGPREDICTIVE MODELING
Single Guest Vie
Propensity
Modeling
Cost of
Acquisition
A/B
Testing
Return on
Marketing
Investment
$0
$50
$100
$150
$200
$250
$300
0 2,000 4,000 6,000 8,000 10,000
Demand Modeling
Forecasting and Pacing
Sales, Base and Churn Monitoring
• Proactively monitor key performance
metrics across all timeseries and
dimensions, autonomously detecting
anomalies against expectations
Insights Generation
• Automate the production of business
insights; providing visibility into why
performance is anomalous and identify,
size the drivers of influence
Customer Analytics
• Integrate disparate data, both internal
and third-party, to understand
customer behaviors and signals
influencing the customer journey
Marketing Effectiveness
• Determine where to spend the next
dollar of marketing investment by
understanding probabilistic
relationships of marketing levers
Pricing – Demand Modeling
• Model demand and revenue curves to
optimize promotional price offerings
and develop pricing strategies
Forecasting / Pacing
• Utilize business drivers – marketing
mix, network, competitive, seasonal
and more – to generate more reliable
predictions of business outcomes

Some European Insurer examples
shared at public conferences
Despite traditional firms generally being
laggards with regards to Data Science &
technology.
Here are a number of examples from other
insurers (shared publicly at conferences):
❖ Vitality (segments + behavioural analytics)
❖ Lloyds Bank (retention analytics/models)
❖ Royal London (predictive underwriting)
❖ Scottish Widows (marketing metrics)
❖ Groupama (data + telematics innovation)
❖ AIG (behavioural biases tests)
❖ Agila (digital analytics + personalisation)
❖ Aviva (intermediary + digital analytics)
25

Increasingly the focus is on
mobile-ﬁrst design principles
NOTIFY A CUSTOMER WHEN
SOMETHING HAPPENS
you need to know what
mattersto a customer, their
latest status and the best
way to alert them, given
the context
RESPOND TO A REQUEST
FOR INFORMATION
you need to know the experience
of a customer and what solutions
you might be able to offer
SHARE AN
EXPERIENCE
to enable social media
sharing you need up-to-
date contact data and all
relevant permissions
COMPLETE A
TRANSACTION
which needs to be
quick and easy, its progress
remembered, and cross-
device migration enabled
COLLABORATE
WITH A CUSTOMER
OR EMPLOYEE
you need to know their
availability, past conversations
and preferences
ENTERTAIN OR
EDUCATE SOMEONE
you need to remember what they
have seen previously in order to
learn about their preferences
CAPTURE OR
CREATE CONTENT
if you enable use of video and/or audio
recording, you need to be able to manage
those ‘Big Data’ challenges
SEVEN TYPES OF DYNAMIC CUSTOMER ENGAGEMENT
However, this will go horribly wrong without accurate dynamic customer data. ‘The Mobile
Mind Shift’4
challenges businesses to design differently for ‘mobile moments’. The authors,
who are Forrester analysts, identify seven types of engagement. Consider how each of
these could go wrong without accurate dynamic customer data:
Source: The Mobile Mind Shift4
With accurate, up-to-date dynamic customer data, organisations
can redesign their processes, thereby meeting increased consumer
expectations of such helpful and engaging experiences.
Could that be a key opportunity for your business?
RESPOND TO A REQU
FOR INFORMATION
you need to know the ex
of a customer and what s
you might be able to offe
COMPLETE A
TRANSACTION
which needs to be
quick and easy, its progress
remembered, and cross-
device migration enabled
CAPTURE OR
CREATE CONTENT
if you enable use of video and/or audio
recording, you need to be able to manage
those ‘Big Data’ challenges
4
The Mobile Mind Shift: Engineer Your Business to Win in the Mobile Moment” by Ted Schadler, Josh Bernoff and Julie Ask (Groundswell Press, 2014)
Source: The Mobile Mind Shift4
26

Following some of the leading
Data Science leaders can help
27
Orlando Machado = Aviva Sanjeevan Bala= Channel 4 Lester Berry = John Lewis
Ryan Den Rooijen = Dyson Graeme McDermott = Addison Lee Martin Squires = Homeserve

Pause to Think:
What might be the
most important
application for you?

Spotting Data Opportunities
(Part 3 of 5)
30

• Video (e.g. accident or property
damage)
• Audio (e.g. customer call)
• Image (e.g. speciﬁed items to insure)
• IoT data (e.g. wearable utilisation)
• Text (e.g. emails/text/social media
interaction)
• Models/Rules (e.g. propensity score
or segment for personalisation)
• Event triggers (e.g. Proxies for
renewal dates, important life events)
Digital Transformation is driving a
wider range of data types to handle
31

Most proﬁtable BigData + Data
Science case studies use internal data
32

Data is still foundational to Data
Science applications
Monica Rogati: The AI Pyramid of Needs. Hacker-noon blog post, 1 Aug 2017
(https://hacker-noon.com/the-ai-hierarchy-of-needs-18f111fcc007, accessed 19/04/2018).33

Old data foundations still matter,
don’t overlook the basics
Predictive + Prescriptive Analytics
Business Intelligence
Data Science & AI
Traditional Analytics
Data Availability & Quality are the foundations for building any part of this house
• Ease of data access (normally through ﬂexible Cloud based solution, AWS et al)
• Freedom to move and transform data (with suitable Data Lake or “sand pit” space available)
• Single Customer View (at least virtual, to enable customer records as basis for analysis
34

Latest EU Data Protection Regulation
has raised the bar (GDPR inc. Article 5)
PROTECTING CONSUMERS
The GDPR has been mentioned a few times already, and we all need to consider its
principle of privacy by design, and by default, in our plans. That means ensuring all uses
of new technology, or major changes, consider their impact on data subjects. Normally
this will be via formal Data Protection Impact Assessments (DPIAs). These must include
due consideration being given to ensuring data accuracy and appropriate fair processing.
2017 research by RMDS confirmed that, for marketers,
GPDR non-compliance had reached the top of customer
data management challenges.
CUSTOMER DATA MANAGEMENT CHALLENGES
2017 – All 2017 – Brands 2017 – Agencies 2016
NON COMPLIANCE
WITH GDPR
29.4%
11.9%
24.6%
35.2%
LEGACY
SYSTEMS
27.7%
37.6%
37.3%
16.2%
POOR DATA
EQUALITY
17.8%
20.3%
16.7%
19.1%
SOURCING
RESPONSIVE DATA
10.8%
11.9%
9.5%
12.4%
NO CHALLENGES
4.8%
9.4%
2.4%
7.6%
DON’T KNOW
4.0%
2.6%
0.8%
4.8%
OTHER
6.9%
4.9%
8.7%
4.8%
Source: Royal Mail Data Services Research 2017
DATA ACCURACY CHALLENGE 2:
THE GDPR AND BEYOND
Since May 2018, the General Data Protection Regulation (GDPR) has been in force in the
UK, enforced by the Information Commissioner’s Office (ICO). UK and European businesses
are coming to terms with what this means. Fines of up to 4% of global annual turnover are
possible under GDPR and recent actions by the ICO indicate a growing willingness to act
where an organisation breaches data protection law.
Consumers have also been woken up to their rights
by a flood of privacy policy and repermissioning emails.
So the risk to brand reputation of being ‘called out’ by
the ICO may be the greatest cost. The ‘Your Data Matters’
campaign by the ICO is also raising public awareness.
A great deal of press coverage on the GDPR has focussed on the higher standard for
evidence of ‘consent’ for data processing. Article 6 of the GDPR, outlining potential legal
bases for data processing, does set a higher bar. Whichever bases for data processing are
them, on average, six per cent of their annual revenues.
DON’T KNOW
40%
30%
20%
10%
0%
5.1%
3.7%
3.3%
5.7%
33.2%
33.7%
20.1%
34.6%
30.6%
23.3%
4.2%
2.6%
16+% 11-15% 6-10% <5% 0%
COST OF POOR-QUALITY CUSTOMER CONTACT DATA AS A PERCENTAGE OF ANNUAL REVENUE
CONCLUSION
In conclusion, dynamic customer data is both a key challenge for today’s businesses and
a potential win-win benefit. Either way, it cannot be ignored. Accurate data has become
the lifeblood of today’s business operations and customer interactions. The need for
timely and accurate dynamic customer data management has never been greater.
‘Dynamic customer data’ is a term reflecting how quickly customer
data is changing, leading to out-of-date data in businesses.
However, there is a positive business case to be made for investing in a solution to
achieve dynamic customer data management. ROI benefits may come from a wide
range of functions including marketing cost savings, improved customer experience
and compliance fine avoidance.
LEGISLATION
(GDPR)
SPEED OF
CHANGE
DEMANDS
ON IT
THREE KEY CHALLENGES
FOR TODAY’S BUSINESSES MAKE ACHIEVING
DATA ACCURACY HARDER:
THREE KEY TRENDS
PROVIDE EXAMPLES OF CUSTOMER ACQUISITION AND
RETENTION OPPORTUNITIES, FROM USE OF DATA:
35

The digitisation of our lives also drives
a faster speed of Data Quality erosion
DATA ACCURACY CHALLENGE 2:
Research from Royal Mail Data Services (RMDS) reveals that
organisations believe that inaccurate customer data costs
them, on average, six per cent of their annual revenues.
DON’T KNOW
40%
30%
20%
10%
0%
5.1%
3.7%
3.3%
5.7%
33.2%
33.7%
20.1%
34.6%
30.6%
23.3%
4.2%
2.6%
16+% 11-15% 6-10% <5% 0%
COST OF POOR-QUALITY CUSTOMER CONTACT DATA AS A PERCENTAGE OF ANNUAL REVENUE
complex interconnected digital economy, that the GiGo theory still applies – that is to say
‘garbage in = garbage out’.
A STEEP CLIMB OVER DATA ACCURACY CHALLENGES
Data gurus3
have offered advice on customer data quality management for over 20 years.
However, many factors make this a more complex challenge than those faced in the past.
The speed of changes in personal data, increased data protection regulation and the
amount of IT change planned all make this feel like a moving target.
Let’s consider those challenges in turn, to see how they may apply to your business...
DATA ACCURACY CHALLENGE 1: ‘SPEED OF CHANGE’
Changes in consumer behaviour, as well as a greater need for up-to-date data, mean
customer data decays at a faster rate than ever before. Akin to use-by dates on food,
data becomes useless if it is not updated to keep track of changing circumstances.
This rate of personal data change has gone unnoticed by too many organisations.
Looking at this in more detail, up to 3,000 changes are made every day to the Royal Mail
Postcode Address File (PAF®
). Couple this with the latest data from the Office of National
Statistics relating to daily life events that also affect customer data accuracy, there are so
many changes occurring on a daily basis for this to easily and quickly lead to out-of-date
addresses, names or inappropriate understanding of needs. The hidden cost to your
business of such inaccurate data reveals itself in many business functions. Inaccurate
addresses cause marketing and product delivery returns, as well as impacting billing
and collections. Inaccurate names deliver poor customer experience, and customers
may leave due to a poor impression of the company. In addition to marketing and
operational costs, there is a greater risk of regulatory fines.
9,590HOUSEHOLDS MOVE
1,496PEOPLE MARRY
810PEOPLE DIVORCE
2,011PEOPLE RETIRE
1,500PEOPLE DIE
DAILY LIFE EVENTS
AFFECTING CUSTOMER
DATA ACCURACY
Source: Office of National Statistics
36

The majority of businesses are not
cleaning their data often enough
pace. More than 15,000 changes are being made daily to people’s personal information,
which equates to 5.6 million changes per year.
The term ‘dynamic customer data’ has been coined to remind us of this complexity. In this
context, data accuracy management is not a ‘once and done’ activity. Organisations cannot
rely on cleaning data on import to a data warehouse, just to then leave the data to rot in
that ‘data graveyard’.
The regularity of data changes, and the need to have up-to-date accurate customer data,
requires rapid data cleaning. Dynamic customer data management is a term covering the
technologies that provide this. These technologies enable organisations to clean new data
and update existing data daily or in near real-time. They allow businesses to keep up with
the consumer life events affecting their customer data.
Worryingly, research from RMDS reveals that almost a third of businesses have no formal
data cleaning process; less than a quarter of the businesses surveyed are implementing
data cleaning on a daily or continuous basis.
FREQUENCY OF DATA CLEANING
DAILY/
CONTINUOUSLY
NO FORMAL
PROCESS
DON’T KNOWMONTHLY QUARTERLY ANNUALLY
40%
30%
20%
10%
0%
18.7%
13.6% 13.1%
9.8% 10.8%
14%
32.7%
37.3%
7.9%
12.4%
7.8%
22%
2016 2017
to increase the need for positive consent.
Beyond data protection,
a number of sectors are
also being challenged
to better safeguard
their customers,
from conduct risk for financial services firms to
increased scrutiny for gambling businesses. As
an example of dynamic customer data impact,
one gambling brand discovered more than 75
duplicate records for one customer. That scale
of error makes safeguarding individuals or
conducting risk management impossible.
Data accuracy is essential for operational processes to protect individuals.
It is also needed for executive reporting and as evidence for regulators.
Failing to address corrupted data accuracy can lead to flawed decisions,
some of which may be critical to keeping your customers.
37

Requirements for a Dynamic
Customer Data Management solution
38
DYNAMIC CUSTOMER DATA
MANAGEMENT SOLUTIONS
REQUIREMENT 1: QUALITY OF REFERENCE DATA
If you are feeling under-prepared, and without a solution to move from your current legacy
systems to this brave new day, there is hope.
Reassuringly, technology solutions are available to meet this need. But, like so much IT,
not all of them are as complete as they appear. So, let’s briefly consider what to look for
in such a solution.
Using external data sources to validate, correct or augment internal customer
data relies on the quality of sources. Check the coverage, quality control,
timeliness and ownership of original data sources. Beware of an impressive
demo, or initial data set, that will degrade over time because of provenance.
Keeping up with dynamic customer data often requires a provider who
owns key data assets. Your due diligence should also require evidence of
their GDPR compliance – for example, the precise permission gained or
how data subjects are kept informed.
REQUIREMENT 2: EASE OF ENTERPRISE-WIDE INTEGRATION
For most businesses, beset with a complex legacy infrastructure, more
on-site systems are not the solution. Look for cloud-based solutions,
combined with easy to use application programming interfaces (APIs).
Such a solution should achieve a lower cost of ownership. It will also
be more flexible to integrate with changing internal systems. Involve
your IT team, and ask providers for details of APIs, as well as any push
notifications from external systems.
REQUIREMENT 3: SECURE AND COMPLIANT SHARING OF DATA
Given one of the key reasons for investing in such a solution is to comply
with the GDPR, this should be a focus. There is no point improving your
compliance with regards to data accuracy, only to breach rules on data
sharing. Look for evidence of security and anonymity. What levels of
DYNAMIC CUSTOMER DATA
MANAGEMENT SOLUTIONS
in such a solution.
encryption are supported for data sent to or from a cloud-based solution?
After initial data provision, how are updates kept anonymous to avoid data
breaches? How secure are their systems and premises?
in such a solution.
encryption are supported for data sent to or from a cloud-based solution?
After initial data provision, how are updates kept anonymous to avoid data
breaches? How secure are their systems and premises?
REQUIREMENT 5: INTELLIGENCE OF CUSTOMER MATCHING RULES
Historically, efforts to achieve SCVs have been hampered by simplistic
matching. Rules or algorithms have not managed to handle the subtle
variations people make to name and address spelling. Look for a
provider who has experience of handling such nuances in addressing.
Ask them about how their rules would handle all the different ways
people address you.
To avoid regulatory fines and take appropriate customer action, achieving
a single version of the truth is essential. Look for solutions capable of
maintaining as near to a real-time single customer view. This is likely to
require several elements of functionality from potential suppliers:
REQUIREMENT 4: REAL-TIME ‘SINGLE CUSTOMER VIEW’ (SCV)
• An initial complete cleanse of all customer data
• Generation, maintenance and sharing of encrypted, unique customer keys
• Integration with all systems across your organisation, to review all new transactions
• Ability to provide these capabilities on demand ‘as a service’
• Encrypted push notifications of changes to be made to related customer records
• Near-real-time updates, notifying key changes in customer data
people address you.
REQUIREMENT 6: A PARTNER YOU CAN TRUST
Just as Rome wasn’t built in a day, so this will not be a ‘once and done’
type solution. Achieving, maintaining and improving the quality of your
customer data is a long-term commitment. Your business will continue
to need such data accuracy for years to come. So, as with all strategic
IT investment, the people matter as much as the technology. Look for a
provider you can trust, one who understands the business of customer
data with an obvious track record of achieving data accuracy and
managing the issues that arise.
Ensure too that they not only understand the implications of the GDPR,
but that they are people you want to work with. A provider who invests
in events and content to educate others is a good sign.
20
people address you.
but that they are people you want to work with. A provider who invests
in events and content to educate others is a good sign.
people address you.
people address you.
people address you.

Pause to Think:
What data is currently not
being used (or cleaned)
in your organisation?

Coding opportunities for all?
(Part 4 of 5)
40

Data Science has also created or made
famous new programming languages
41

It has become popular to learn to
code and many guides are available
42

There is confusion over the term, other
software pretends to be Data Science
43

A key advantage of Data Science
coding languages is ecosystems
44

Students leave university coding in
these languages + communities
45

Pause to Think:
Do you or one of your
team need to learn to
code in R or Python?

Getting ready for Data Science
(Part 5 of 5)
47

5 keys from experience of helping
businesses implement Data Science
48
1. Access to Data
2. Right Tools for the job
3. Domain knowledge available
4. IT team in the loop
5. Clarity of goals/priorities

Ensuring sufﬁcient access to data
for Data Science team
49

Don’t be confused by different
software labels
50
the industry, it is vital that insurers quickly start to accelerate the acquisition of value from their analytics program.
BUSINESS CAPABILITIES FOR INSURERS
Answers Insurers Are Seeking
Business intelligence and analytics are becoming increasingly vital to every part of the insurance business. Insurers
need capabilities that address a wide variety of questions across marketing, sales, and service as well as enterprise
operations. The general types of questions raised are illustrated in Figure 1. At a high level, insurers want to explore
questions like the ones above the orange boxes: How do we gain new insights from historical data? What are our
new opportunities? At the next level, they are asking more specific questions, like those inside the orange boxes:
What happened? Why is it happening? What can we do about it?
There are a variety of technology tools and approaches to address these questions. They generally fall under the
categories of business intelligence, advanced analytics, and emerging analytics, and these technologies can be
applied to answer the types of questions posed in each of the sections.
Figure 1. The BI and Analytics Spectrum for Insurers
Source: Strategy Meets Action 2016
As can be seen from the diagram, the questions to be answered range from the traditional, more operational
types of issues to more complex and differentiating insights and actions. On the far left, analysis of historical
data enables reporting on the state of the business (What happened? What is happening now? Where is the
problem?). In addition, historical data (both internal and external) can help with diagnostics on specific problems
(Why is it happening? What if it continues?). Towards the middle and the right of the diagram, more complex
and forward-looking analytics can be applied to understand how insurers can identify, predict, and capitalize on
new opportunities, and ultimately, through emerging analytics, move to human augmentation and automated
decisioning. It should be noted that big data is an overlay onto this diagram, providing a set of approaches and
technologies to answer these questions when the volume, variety, and velocity of the data cannot be addressed in
a timely manner by traditional analytics.
How do we gain new insights
from historical data?
BUSINESS INTELLIGENCE
What are our new
opportunities?
How do we capitalize on new
opportunities?
How do we leverage
human intelligence?
ADVANCED
ANALYTICS
EMERGING
ANALYTICS
Analytic
Collab-
oration
Predictive
Modules
Predictive
Analytics
Data & Text
Mining
Advanced
Statistical
Analysis
ScenariosAnalysis
Ad-hoc
Queries
Dashboards
&
Scorecards
Reporting
Cognitive
Computing
What can we do about it?What is likely to happen?
What if it
continues?
Why is it happening?
Where
is the
problem?
What is
happening?
What
happened?

The critical need to manage
domain knowledge
51
Tips to develop domain knowledge ready
for your Data Scientists:
1. “Know your numbers” = develop
commercial awareness in all analysts
2. Use team meetings to ensure clarity on
big picture & the “why” of work
3. Get out & about, including potentially
shadowing or job swaps (e.g. Strategy)

Need to manage relationship with
IT carefully (like a marriage)
52
The best home for Data Science?
Benefits of “end user computing”, but the
need for process to “promote to live”
Centralised verses Decentralised teams?
Central Design Authority?
Greater data access & the need for a
“playpen” or “data lake” for experiments
Data and Goliath
If de-centralised, where does your
data function and team members sit?
These interesting results
reflect both a significant
change from last year and
a concern I have about IT
ownership. Last year we
saw that most decentralised
teams were scattered across
businesses (46%), which can be
very inefficient.
However, the picture has now
changed to most decentralised
data teams residing in parts of
IT. That concerns me. When
data governance or analytics
are left to IT, business leaders
can tend to obfuscate their
responsibilities , and analytics
can be reduced to inflexible
projects.
I have seen much more
success with data expertise
residing within business and
with analytics functions
sitting within Marketing or
Operations. I hope this is not a
further step on the road where
Data Science is in danger of
reducing analytics to coding.
ឣ IT 51%
ឣ Scattered across the business 27%
ឣ Marketing 15%
ឣ Operations 7%
51% 27% 15% 7%
As a result of
decentralisation, 51%
of our respondents’
data teams are now
located within the IT
department.
Key Findings:

How to achieve clarity of goals &
why that is needed
53
Don’t hire a Data Scientist without
being clear on how they could help
Don’t hire a Data Scientist without
being ready to set them clear goals
Do work with them to prioritise
business challenges & opportunities

Pause to Think:
Are you ready to start
using Data Science?
If not, what are your gaps?

Action-orientated education:
the biggest predictor of value…
55
One thing I will do
differently as a result of
today is…
01 All around us, but let’s get clearer
What is it?
02 Some exciting & concerning applications
How is it being used?
04 How could you learn the coding needed?
Coding Opportunities
03 How can you spot new data & uses?
Data Opportunities
05 What is needed to get started?
Getting ready

Data Science is about a learning
journey, here’s a blog to help you
56

Other resources to help your CPD
EDISON Data Science Framework:
Part 1. Data Science Competence Framework (CF-DS)
Release 2
Project acronym: EDISON
Project full title: Education for Data Intensive Science to Open New science frontiers
Grant agreement no.: 675419
Due Date
Actual Date 3 July 2017
Document Author/s Yuri Demchenko, Adam Belloum, Tomasz Wiktorski
Version Release 2, v0.8
Dissemination level PU
Status Working document, request for comments
Document approved by
This work is licensed under the Creative Commons Attribution 4.0 International License.
To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/
57

Checking in on Baron von
Munchhausen…
58
Goals:
Did you get what you
needed from this
masterclass?

customerinsightleader.com
laughlinconsultancy.com
@LaughlinPaul
linkedin.com/in/paullaughlin
paul.laughlin@southwales.ac.uk
+44 (0)7446 958061
Any questions?

Everyday Data Science

Recommended

Recommended

More Related Content

Similar to Everyday Data Science

Similar to Everyday Data Science (20)

More from Paul Laughlin

More from Paul Laughlin (20)

Recently uploaded

Recently uploaded (20)

Everyday Data Science