BIG DATA - lifeconex.com€¦ · Big Data – A Journey in Life Sciences & Healthcare 11 So instead of defining big data as a number or a size, it is more interesting and relevant

BIG DATAA JOURNEY IN LIFE SCIENCES & HEALTHCARE

Big Data – A Journey in Life Sciences & Healthcare2

BIG DATAINSIDE

01 Introduction: The Fourth Industrial Revolution 4

The Fourth Industrial Revolution is a new chapter

in human development.

02 Big Data & Advanced Analytics 8

The objective to achieve with big data is relatively

simple: your decisions and actions should be based

more on data analysis and less on intuition.

03 Case Study: Data-Enhanced Products 12

These new technologies are changing how

organizations perceive and manage their assets.

04 Conclusion: Democratizing Data 18

Everyone is going to have to become smarter.

Don’t fight the robots.

3Big Data – A Journey in Life Sciences & Healthcare

04

Start by identifying your critical areas.

This requires sitting down with

different managers and subject matter

experts across your organization.

Processed knowledge

is wisdom.

Advanced analytics techniques such as

predictive and prescriptive analytics are the next

steps that help convert data into knowledge.

03

02


INTRODUCTION:

THE FOURTH INDUSTRIAL REVOLUTION

01

Big Data – A Journey in Life Sciences & Healthcare 5

The Fourth Industrial Revolution is a new

chapter in human development driven by

the increasing availability and interaction of

a set of extraordinary technologies, building

on three previous technological revolutions.1

An example of this technology can be seen

in smart factories that are augmented with

web connectivity and can visualize the entire

supply chain in order to make decisions on

their own.

The question for all industries and companies, without

exception, is no longer “Am I going to be disrupted?”

but rather “When is disruption coming, what form will it

take and how will it affect me and my organization?’’2

These new technologies are impacting and re-defining

all governments, economies and industries. They have

the potential to connect billions of people to the web,

drastically improve the efficiency of humanity and

help regenerate the natural environment – potentially

even reversing the damage caused by previous

industrial revolutions.

“ It’s the decade of data and

that’s where the revolution

will come from.”

Alex Pentland

But there are also tremendous risks. Organizations could

be unwilling to adapt to these new technologies and

governments could fail to employ or regulate them

properly. Putting it in the perspective of Life Sciences and

Healthcare products and their associated logistics and

supply chain – we have a tremendous responsibility

together as partners to maintain our processes. The

products that you manufacture and that we transport will

continue to impact all of humanity. We must stay

committed to utilizing all data available to mitigate risk

for the end receivers – the patients.


Figure 1: Fourth Industrial Revolution

Source: Forbes

1st 3rd2nd 4thMechanization, water power, steam power

Mass production, assembly line, electricity

Computer and automation

Cyber Physical Systems

Impact felt already 2015-2017 2018-2020

Figure 2: Time to Impact Industries’ Business Models

Source: World Economic Forum


We are being bombarded by new entrants offering new business models, disrupting and influencing the

strategies of established enterprises who, as a result, are rapidly transforming their businesses. The one thing

all participants have in common, whether at industry level, company, or individual professionals - is that data is

the driving force behind all new developments.

What are the cold chain trends related to data that you should pay attention to in 2018 and beyond? The aim

of this article is to provide the reader with a projection of these technological concepts by discussing practical

examples of how we are partnering together with our customers today to successfully draw out insights for

decision-making from our data.

n Rising geopolitical volatility

n Mobile internet and

cloud technology

n Processing power, Big Data

n Sharing economy,

crowdsourcing

n Middle class in

emerging markets

n Rapid urbanization

n Changing nature of work,

flexible work

n Climate change, natural

resources

n New energy supplies

and technologies

n The Internet of Things

n Advanced manufacturing

and 3D printing

n Longevity and ageing societies

n New customer ethics,

privacy issues

n Women’s economic

power, aspirations

n Robotics, autonomous

transport

n Artificial intelligence

n Adv. materials, biotechnology

02


BIG DATA AND ADVANCED ANALYTICS: A CLOSER LOOK


One of our top global pharmaceutical

customers recently requested to align with

DHL and put forth a strategy that combines

all of the forwarder and shipper data that

we have at our disposal, including but not

limited to: near product temperature sensor

readings, ambient temperature sensor

readings, logistics milestones, shipping

performance, lane risk assessments, issues

analysis, and CAPA (Corrective and

preventative action) analysis – with the goal

of creating an end-to-end Risk Based

Approach.

The objective to achieve with big data is relatively simple:

your decisions and actions should be based more on data

analysis and less on intuition. This enables you to reach

your goal – which is the ability to transform data into

information – and information into knowledge – so that

you can optimize the process of decision-making within

your business.

As a term, big data still has no clear definition. For some, a

dataset over a terabyte is big data – for others, it might be

a million rows, and others still may have smaller datasets

that are changing multiple times per second.

“ Processed data is information.

Processed information is knowledge.

Processed knowledge is Wisdom.”

Ankala V. Subbarao

PR

EDIC

TIV

EANALYTICS

REGRESSIONSTATISTICAL

BA

SIS

DET

ERM

INE

USEDM

OD

ELLI

NG

DU

RA

TIO

N

BEHAVIOUR

DEPENDENT

CONSUMER

PORTFOLIO

LEARNING

AD

APT

IVE

DESCRIPTIVETAXES

CLASSIFICATION

AN

ALY

ZE

NEW

APPLICATIONS

CRM

WELL

FRAUD

PROCESSP

RO

CES

S

BASED

MA

NA

GEM

ENT

CR

LINEAR

HAZARD

PATTERNS

EVEN

TS

FUNCTIONS

METHOD

RISK

HELP INDEPENDENT

TRAINING

STA

TIO

NA

RY

MANY

SER

IES

APPLIED

SET

ETC

MET

HO

DS

VECTOR

PRED

ICTI

NG

FUNCTION

DIS

CR

ETE

PERFORMANCEM

AK

E

STAGE

SUPPORT

NEURAL

APPLICATION

OR

DER

MEDICAL

LEV

EL

SEE

MA

CH

INECONDITIONAL

NUMBER

EXAMPLES

INFO

RMAT

ION

LIKE

OFT

EN

COLLECTION

RELATIONSHIPS

TYPESNEGATIVE

LOGIT

HEALTH

DIFFERENT

EXA

MP

LE

TIME

USI

NG

SURVIVAL

SURV

IVA

L

CA

SES

LIKELIHOOD

SYSTEMS

MO

DEL

S

NORMAL

FUTURE

SERVICES

OBSERVATIONS

SERV

ICE

DATA

DETECTION

PREDICT

EVENTS

MULTINOMINAL

SAMPLE

AN

ALY

SIS

CU

STO

MER

STE

CH

NIQ

UES

VA

RIA

BLE

SN

ETW

ORK

S

DIS

TRIB

UTI

ON

PROBIT

TOOLS

ALS

O PR

OD

UC

T

ONE

RESPONSE

POSITIVE

RULES

ESTIMATION

RA

ND

OM

ASSUMPTIONS

PEO

PLE

IDENTIFY

MARKETING

FOC

US POTENTIAL

LOGISTICS

CONTINUOUS

1010 Big Data – A Journey in Life Sciences & Healthcare


So instead of defining big data as a number or a size, it is

more interesting and relevant to define it in terms of value.

After all, your data is only as good as the insights that you

are able to gain from it. The speed with which you are able

to translate those insights into recommended actions, and

communicate them to the relevant decision-makers within

your organization is becoming more and more critical.

In the previously mentioned example, we combined many

different types of data to form insights that helped us to

create a risk management process that we could replicate

and scale throughout the customer’s entire network. Let’s

take a closer look at the details:

Risk Identification: The customer sent DHL their

internal temperature data logger data on specified

trade lanes in which they had experienced temperature

deviations.

Risk Assessment: DHL compared the internal data

logger data with our external data logger data (which we

put on the outside of shipments) to examine the

temperature curves.

Risk Analysis: DHL matched temperature spikes to

logistics milestones, and determined that all of the

deviations occurred while in the airport to airport

environment. This was during the time the various airlines

maintained possession of the cargo. DHL further examined

the behavior of temperature data and trends related to

lane risk assessment variables associated with the airlines.

Risk Control: DHL and the customer jointly determined

what the level of acceptable risk would be within the

SOPs. DHL recommended to the customer that we only

utilize DHL Life Science & Healthcare Preferred airlines

and/or airlines with pharmaceutical services within our

processes. Each SOP was required to have at least 3

contingency routings. Options were selected that had a

high frequency of flight options to avoid problems with

uplift capacity.

Risk Acceptance: SOPs within scope of this pilot were

updated and reviewed by all parties. The customer accepted

an overall lower risk by selecting airlines that have a higher

proficiency in handling temperature-sensitive products.

VALUE TO CUSTOMER:

Risk Management: Utilization of data-driven

insights to create a scientific approach for

strategic decision making in risk management.

Data Engineering: Custom data collection,

reporting and analytics dashboards implemented

to identify and visualize immediate tactical

improvements for consideration.

Options Analysis: Identification of higher quality

airline service levels for optimal cost.

Risk/Benefit Analysis: Examine direct vs. indirect

costs and removal of non-value added services.

Risk Reduction: Decrease quality issues

in distribution.

Residual Risk Evaluation: Determine additional

improvements (seasonality related to service

for example).

Communication/Synchronization: Alignment

on terminology between DHL and the customer

(Label/Storage vs product/transportation

for example).

“ Our customers want these risk

mitigation processes supported by our

tools to be integrated with their

shipping strategy, with the overarching

goal to reduce temperature excursions.”

03


CASE STUDY:

DATA-ENHANCED PRODUCTS


BACKGROUND

Sensors and telemetry embedded in physical objects, which

are connected to the internet, are collectively described as

the Internet of Things (IoT). These new technologies are

changing how organizations perceive and manage their

assets, and have become a cornerstone of integrated and

embedded systems feedback and control. Analysis

provided by sensors, that are placed on assets, enables

their constant monitoring and proactive maintenance and

in doing so maximizes their utilization. The focus has thus

shifted to using performance benchmarks (based on data

and monitored through algorithms) that can highlight

when a piece of equipment is moving outside of its normal

operating parameters.

The logistics and transportation sector generates

enormous amounts of data that are exploited using big

data technologies. In shipping one container there are

more than 200 interactions and 30 people involved.3

Companies are combining geo-location tools with

advanced analytics to establish more efficient logistics

routes. Access to the real-time status of their networks

means improved efficiency is now a reality. Tracking

products is more accurate, more reliable and predictions

can be made about the demand for a service – so that

they can move forward to the market and manage their

workloads more efficiently.

“ Hiding within those mounds of data is

knowledge that could change the life

of a patient, or change the world.”

Atul Butte


As an industry leader with extensive experience in good

distribution practice and regulatory compliance, DHL has

gained a unique perspective on the issues posed by large

amounts of complex data. Since 2012, millions of data

points have been collected from an expanding array of

sensors that monitor ambient temperatures captured

throughout the DHL global shipping network.

A large knowledge database is maintained where every

shipment has a process management record number

associated with temperature and location data. This

database has become too large to extract insights through

human means, and that’s where machine learning comes

into play.

Machine learning has a number of problems that it has the

potential to solve and can typically be grouped into the

following categories: classification, clustering, regression,

and optimization. Classifiers categorize data, typically into

either concepts or collections that utilize the notion of

resemblance. Clustering looks to identify a pattern in the

data. Regression typically extrapolates the trends present in

the historical data to help the system make a prediction

based on those trends. Optimization, the end goal so to

speak, is then all about improving your model via

evaluation and iteration. 4

10

7 8 9 10 11

0

20

30

40

50

Tim

e (

hrs

)

Reading Temperature2017

2. Export Forwarder

3. Carrier

4. Transit

5. Destination Airport

High Risk (40+)

1. Customer Warehouse

SOLUTION

DHL has developed a risk model, based on historical

and real-time data, which assigns a process deviation

estimate for each step in a shipping procedure.

This allows logisticians to evaluate the root causes and

critical areas of risk in the logistics process (e.g. specific

airport, section of route, packaging type, time of year,

airline service, etc.).

This type of analysis is especially critical when shipping

high-value, life-saving products.

Where have we experienced higher risk?


2

0

4

6

8

Trad

elan

e 1

Trad

elan

e 2

Trad

elan

e 3

Trad

elan

e 4

Trad

elan

e 5

Trad

elan

e 6

Trad

elan

e 7

Trad

elan

e 8

Trad

elan

e 9

Trad

elan

e 10

Trad

elan

e 11

Trad

elan

e 12

Trad

elan

e 13

Trad

elan

e 14

Trad

elan

e 15

Trad

elan

e 16

Airline A

Airline B

Airline C

Airline D

Airline E

Airline F

Airline G

Airline H

Airline I

58%

4

0

8

12

16

5

0

10

15

20

Te

mp

era

ture

(C

)

Tim

e (

hrs

)

ToR

AVG Temperature (OR)

16.75

9.75

19.5

13.19

3.5

13.65

Where should we focus our efforts to reduce risk?

Which entities are causing the most risk?


Advanced analytics techniques such as predictive and

prescriptive analytics are the next steps that help convert

data into knowledge. This typically goes far beyond having

just an Enterprise Resource Planning (ERP) system where

you study the company data in isolation. More often than

not, companies don’t have, or don’t make the effort to

get, third-party data to help them interpret their internal

data. Instead they are influenced by preconceived ideas or

beliefs that may or may not be entirely accurate.5

DHL is focusing on combining data generated from their

risk model with data sets shared by partners and customers

to discover and extract patterns of information that were

previously unknown. Once scenarios are generated from

all possible process actions found in the database, risk

values are assigned using a probability of failure formula.

Once you have reached this point, you are able to utilize

the entirety of your data to draw insights, make decisions,

and continuously improve.

DESCRIPTIVE

n Humans gain insight and

take action

PREDICTIVE

n Humans gain insight and take

action

n Interactive user experience

n Democratization of access

PRESCRIPTIVE

n Humans gain insight and

take action

n Highly diverse textual and

numerical sources

n Machines gain insight,

make recommendations

and take action

Source: DHL Temperature Management Solutions

BENEFITS

Proactively make data-driven recommendations

to customers with new trade lane verification

and prioritization for new product introduction

or facility introduction or change.

Collaborate with customers to prioritize trade

lanes and ensure rapid deployment of product

into market.

Provide benchmarking capabilities to compare

performance relative to industry averages.

Evaluate opportunity to reduce costs by utilizing

lower-risk lanes, less costly packaging, and/or

lower cost transportation modes.

Evaluate companies’ performance/risk during M&A.

Step 01:

“Just in time”

Step 02:

“Near real-time”

Step 03:

“Real-time”


“ In 2018 we are likely to see these capabilities taken further with the inclusion of a greater number of external data sets” states Jasmine Miller, Data Engineer, DHL. In effect, historic and real-time data is combined to say “this is the right thing to do right now, in your situation. The end goal is of course to focus on identifying high-risk components of shipping procedures, so that logistics specialists may reduce temperature excursions in transit and provide recommendations for continuous improvement in the case of a process deviation.”

04


CONCLUSION:

DEMOCRATIZINGDATA


Everyone is going to have to become

smarter. Don’t fight the robots. Instead, fall

in love with workflow automation. Familiar

problems can be handled by bots. Highly

unique tasks are where humans come into

play. Use technology to free up people’s

attention as much as possible to work on

complex problems. People with

self-confidence in their own skill set can

then start to tackle the changes that are

occurring in the fourth industrial revolution.

Start by identifying your critical areas. This requires sitting

down with different managers and subject matter experts

across your organization. Decide what areas within any of

your ‘decision-making processes’ can be improved, but also

be realistic about it.

Consider a small use case which you want to tackle and

determine its feasibility. You need to determine if you have

the data to support the solution you are looking to solve. To

create possible scenarios by which you can make instant

recommendations, you need data which will be able to:

anticipate possible outcomes (what will happen), the reason

for them occurring (why it will happen), and the time of

their occurrence (when it will happen).

“ To remain competitive, both

companies and countries must be at

the frontier of innovation in all its

forms, which means that strategies

which primarily focus on reducing

costs will be less effective than those

which are based on offering products

and services in more innovative ways.”

Klaus Schwab


Your data is the foundation of any type of analytics

solution. It must be clean, accurate and provide value. To

help with this, you might be required to invest in a

platform strategy (which is both profitable and disruptive).

To help narrow down the many options out there, you

should ask what the goal of the platform is:

Do you want to ingest a large variety and volume of

structured and unstructured data?

Are you going to combine your own data and other

sources of information with advanced techniques

such as machine learning?

Do you want customers to have access to these

insights via a platform – or is it only an internal tool

used for continuous improvement?

Is your ambition to have something in place which

does all of these things?

These types of questions will guide you in determining

which solution platform is best for your business. Keep in

mind that any type of advanced analytics solution is not

going to be perfect, and you must be conscious of the

issues you might encounter when investing. For instance,

companies have encountered issues where clients want to

understand what models are driving the solution. Most

machine learning models are advanced mathematical

algorithms – if you can’t explain the model to your client,

they might not be interested in utilizing your solution.

You must also consider that your solution is constantly

learning from whatever data you are training it on. Unlike

a human, which can translate skills and knowledge to

multiple domains and problems, your models are trained

on the data you provide. If there is a scenario in which you

haven’t trained the platform, it might not be able to

identify the problem. Self-learning systems with data

mining and pattern recognition capabilities that can learn

from a constant incoming stream of information are the

goal – but we will have to work together and combine our

insights to achieve such a goal. There will come a time,

sooner rather than later, that self-adaptive decision

services can recommend optimized shipment parameters

upon receiving a shipment service request – so we should

all be prepared for that.

To summarize, the core of these macro-trends is about us

as human beings and how we embrace all of this change.

Leveraging big data will enable faster and better decision-

making in a wide range of industries and applications.

Automated decision-making can allow businesses to

provide real-time services and support every conceivable

customer interaction. Imagine what your managers could

do if they were able to orchestrate the advanced analytics

and automated behaviors of different applications

discussed in this article. They would be able to:

Build processes that are unconstrained by developer

availability and time constraints

Automatically trigger actions across departments

based on the most up-to-date data

Focus less on technology, and more on their roles

within the organization

Democratizing data in these ways will empower the

workforce to improve processes that are typically routine

and repetitive. This will free your teams from low-value

work so that they can develop skills for higher-value work

that require complex problem solving and critical thinking

– both of which are in high demand.

In our journey to where we are today, we have learned

that it’s important to explore and experiment to succeed.

Data science projects are challenging and have many

unknowns which only become clear as you get down into

the details.

The evidence is clear: We’re in the midst of a data

explosion driven by the increased ability to collect, store

and analyze data. As we go deeper into the reality of the

fourth industrial revolution, big data will play an

increasingly crucial role. The only limitation to leveraging

the power of data is your own imagination.


ABOUT THE AUTHOR:

Gordon Johnson, Global Head of

Optimization, Temperature Management

Solutions, leads DHL’s Cold Chain

Optimization team, a consulting group

which designs transportation solutions

that enhance customer satisfaction and

success in the Life Sciences and

Healthcare sector.

Gordon began his career with DHL in 2004 and was

one of the founding members of LifeConEx – a joint

venture start-up formed by DHL and Lufthansa

Cargo. He has held various roles throughout his

career at DHL, primarily focused in corporate strategy

and innovation, product development, and risk

management. He is currently leading the creation

and implementation of advanced analytics tools

designed to increase compliance with Good

Distribution Practice and reduce shipping risk for Life

Sciences & Healthcare companies.

Gordon earned his degree in molecular biology and

biotechnology from Florida Atlantic University in

2003, and holds an executive certification in

strategy and innovation from the MIT Sloan

School of Management.

REFERENCES1 Klaus Schwab, “Shaping The Fourth Industrial Revolution,”

(2018): 23.

2 Klaus Schwab, “The Fourth Industrial Revolution,”

(2016): 13.

3 John Churchill, “Maersk and IBM target one of trade’s biggest barriers”

(2017):

https://www.maersk.com/stories/maersk-and-ibm-target-one-of-

trades-biggest-barriers

4 Mark Skilton & Felix Hovsepian, “The 4th Industrial Revolution:

Responding to the Impact of Artificial Intelligence on Business”

(2018): 149-150.

5 Jorn Lyseggen, “Outside Insight – Navigating a World Drowning in

Data” (2017): 38-39.


VOICE OF CUSTOMER: HERE ARE A FEW REQUESTS THAT OUR CUSTOMERS HAVE BEEN ASKING US TO EVALUATE IN COMBINING OUR DATA TO POSITIVELY IMPACT THE COLD CHAIN

How can we expand on data collected to better protect against painful areas such as the tarmac? For example,

can we add temperature extremes to the lane risk assessment ambient temperature? As the average temperature

is often the best-case scenario – we want to also see what the temperature can be in worst-case scenarios.

Can we have trade lane risk assessments be generated at the time of quotation? We see this as providing a good

balance between cost and risk when evaluating decision-making regarding packaging, airline pharma services,

and other variables.

Can you help us to explore further the aforementioned cost of quality in these various scenarios? Are you (DHL)

able to provide the financial impact for us, so that we can decide if we want to implement a process change, or

accept the costs of non-compliance?

Can you help us to utilize data to show what packaging types have the best performance? For example, is it less

risky to ship via older generation Envirotainers that use dry ice? Or passive packaging? Please use your historical

data to assess what is the best passive packaging type to use on lane x at x time of year with x carrier through x

transit point on this route. And please compare it against all of my contingency routings and let’s see what has

the least process deviations and lowest costs.

Can you integrate our cold chain IT management system with our supply chain disruption and security IT systems?

We need them to share data back and forth and we would like a single sign-on. By the way, can we set up an API

with you to share our internal temperature data logger data automatically, as to overlay on top of your logistics

milestones and pinpoint process deviations, automatically generating and measuring CAPAs?

May we have portal access to anonymized data from your entire data set? We’d like to use this data to “predict

and prescribe” the best routes, carriers, and packaging types based on performance - so that we can choose the

best options from the outset.


Temperature

Management Solutions

DHL Global Forwarding

For further information

contact our supply

chain experts:

www.dhl.com/contact-us

June 2018

BIG DATA - lifeconex.com€¦ · Big Data – A Journey in Life Sciences & Healthcare 11 So instead of defining big data as a number or a size, it is more interesting and relevant

Documents