Operations Practice

Enabling a digital and analytics

transformation in heavy-industry
manufacturing
To make the most of new digital approaches in manufacturing, heavy-industry players
need to put some critical enablers in place.

by Milan Korbel, Stuart Sim, Ken Somers, and Joris van Niel

© Getty Images

December 2019
 In earlier articles, we have shown how digital Critically, projects also require people whose
and advanced analytics (DnA) technologies have skills bridge these different groups. Data
significant potential to improve manufacturing engineers develop more efficient information
operations in heavy industry. We have explained that technology systems such as databases, fast
to capture that potential at scale, companies need data processing, or new and more reliable data
a systematic approach to identify, develop, and roll sources. Data scientists use those systems to
out the technologies and approaches that offer the unlock new insights or knowledge from the data
most value to them. And we have shown how the by developing analytical techniques and efficient
best approach for a given organization will depend algorithms. Translators frame business problems
on its current level of digital maturity, its goals, and in a way that digital specialists understand, and
on the nature and distribution of those opportunities. use their domain knowledge to evaluate and
continuously refine the resulting DnA solutions
In this article we turn our attention to four critical (Exhibit 1).
enablers: talent, data, technology, and agile delivery.
Companies need these enablers in some form With the exception of tactical interventions,
for every digital initiative they run, and together where most of the necessary talent may be
they comprise an engine to power any heavy provided by the external supplier contracted
manufacturer’s successful DnA transformation. to deliver the project, talent can be a critical
Carefully thinking through the development of roadblock in digital projects. And it won’t be
digital muscles helps an organization avoid the solved by the market alone. With a wave of
need to repeat work or reinvent the wheel, aids digitization underway across industries, skilled
the efficient use of resources, and promotes the data scientists and other technical specialists are
adoption of standardized approaches that are easier in short supply globally. Furthermore, becoming
to scale, replicate, and sustain. much more efficient in those roles requires
an in-depth understanding of manufacturing
Let’s look at each in turn: processes and other company-specific
knowledge, alongside digital and analytics
expertise.
Talent
Digital and analytics projects are skill-intensive While some external hiring is almost always
activities. Recent advances in software and necessary to fill capability gaps and kick-start
hardware have done much to improve the digital programs, many critical new skills are
accessibility and usability of new technologies, but best developed in-house (Exhibit 2). Homegrown
their successful application still requires people who capability-building efforts can’t meet an
understand the capabilities and limitations of digital organization’s demand for PhD-trained data
approaches, and who know how to get the best out scientists, but they can produce large numbers
of the available tools. of skilled and competent digital practitioners,
who are essential for the application of new tools
In heavy industrial manufacturing, digital projects at scale. Heavy-industry manufacturers are well
require new capabilities—and often new roles—in positioned in this regard, since their existing
three areas of the business. They need people with workforce is already technology savvy. For them,
expertise in the company’s products and production the shift to digital is a natural step: for example,
processes. They need technology specialists with an automation engineer already understands the
expertise in areas such as software development, building blocks of robotics technology.
robotics, and automation. And they need digital
specialists who can run agile projects or design an Companies embarking on large-scale
effective user experience. transformation programs (whether focused

2 Enabling a heavy-industry digital and analytics transformation

Web 2019
Enabling a digital and analytics transformation in heavy-industry manufacturing
Exhibit 1 of 4

Digital
Exhibit 1 projects require new roles spanning three core business functions.
Digital projects require new roles spanning three core business functions.

Business-domain expert

Product owner Functional expert

Ensure future data
requirements and Translate business needs
processing are Business into digital language (eg by
robust and complete defining data requirements)
Data Translator
engineer
Software/UX¹ developer Business UX/UI2 designer
leaders

Robotics &
automation Data
IT Digital Scrum master
expert scientist

Data architect
Agile coach

Develop insights leveraging new sources

of data and best-in-class algorithms

1. User experience
2. User interface

Source: McKinsey

or enterprise-wide) may set up a dedicated digital digital projects does much to eliminate fear of new
academy to develop skills at all levels of the approaches, while also making their benefits clear.
organization. These academies provide online, Staff can usually see immediately how digital tools
classroom-based, and experiential training tailored can improve their working lives.
to the objectives of the program and the needs of
different stakeholder groups. Senior managers
may be provided with a broad overview of digital Data
approaches, for example, while translators and Most heavy-industry players already have the
digital specialists dive deeper into specific tools. majority of the data they need. But an important
new requirement in digital projects is to extract that
In-house academies also make it easier to deliver data from silos across the organization, and bring
specific training on a just-in-time basis, so staff can it together in a unified, accessible form. Executives
apply new skills immediately on real projects. In line sometimes worry that a digital transformation will
with general best practices in capability building, require them to rip out and replace their existing
learning should then be reinforced with continual data infrastructure. Many have uncomfortable
coaching and mentoring in the workplace. Working memories of the large-scale IT projects of the past,
this way, academies don’t just build skills—they which were expensive, complex, and prone to failure.
also change mind-sets. Hands-on involvement in

Enabling a heavy-industry digital and analytics transformation 3

 Web 2019
Enabling a digital and analytics transformation in heavy-industry manufacturing
Exhibit 2 of 4
Exhibit 2
Theright
The rightDnA
DnAtransformation
transformationtalent
talent pool
pool combines
combines in-house
in-house capability
capability building
building and
and hiring for specific specialist
hiring for specific specialist roles. roles.
Main source
Current job DnA¹ role of talent
Senior Management team DnA In-house
management champion capability
General managers building
Operational Production and
management technology managers DnA In-house
aware capability
Functional team leaders building

Production Maintenance &

In-house
units production team leaders Translator capability
Section managers building

Local IT
Data In-house
Quality & process scientist capability
technologists building

Central maintenance Data In-house

engineer capability
building
Functions Continuous-improvement
project resources IT Hiring
architect external
Central supply chain, talent
marketing & sales, and Product owner, In-house
technology functions Agile coach, capability
Non-technical functions Scrum master building
(HR, finance, procurement) UX/UI2 Hiring
designer, external
External Externally hired talent Software talent
talent engineer

1. Digital and analytics

2. User experience/user interface

Today’s reality is different. Modern data platforms shift reports may exist only on paper, for example,
or data-integration systems can be set up as a and non-networked or analog sensors may need to
new layer that sits above existing systems and be upgraded or replaced. The important rule, which
interacts flexibly with them. This may involve the avoids failures and minimizes costs, is to gradually
use of a data lake, fed by streams of data from a integrate only the data that is needed for ongoing
variety of operational technology (OT) sources, DnA projects.
such as sensor networks, process-control
systems, or other manufacturing-automation Once companies understand their data, they need
equipment—or from higher-level IT sources, such to manage it. That requires robust governance
as enterprise resource planning systems. polices and processes along with rigorous
standardization—using consistent naming
The process isn’t entirely plug-and-play: some conventions and a single “data dictionary” to
data sources may require additional action prior ensure future digital projects can identify the data
to integration. Historical maintenance records or they need (Exhibit 3).

4 Enabling a heavy-industry digital and analytics transformation

Web 2019
Enabling a digital and analytics transformation in heavy-industry manufacturing
Exhibit 3 of 4

Data isExhibit 3
a critical resource that requires effective governance.
Data is a critical resource that requires effective governance.

2 Data
accountability

3
Data
Policies, prioritization

6 standards, &
processes
8
Data
privacy &
1 security
Data vision
and goals
5 Data
quality
7 Data
retention

and
controls

4 Metadata
and data
lineage

1 Data vision and goals to set the direction of the program and focus investment

2 Data accountability to define who is responsible and ensure ownership

3 Data prioritization to focus effort on the data that matters the most

4 Metadata and data lineage to describe the data and where it comes from

5 Data quality and controls to ensure data is clean and data risks are managed proactively

6 Policies, standards, & processes to describe how governance should work

7 Data retention to control the amount of data needed to process and manage

8 Data privacy & security to protect data assets and ensure sensitive information is safeguarded

Enabling a heavy-industry digital and analytics transformation 5

 Data quality is a significant, and often applications, inherent measurement errors can
underestimated, challenge that should be be significant, resulting in uncertainty that may be
addressed early in any digital effort. This will significantly larger than the typical 1 to 3 percent
require processes to identify and manage performance improvements unlocked by DnA
missing or erroneous data. Companies also need solutions.
to ensure they understand the characteristics
of the data they have: the frequency and timing Companies also need a systematic process to
of measurements can have a significant impact identify potential risks and define appropriate
on the ability of a dataset to usefully represent mitigation actions and improvements (Exhibit 4).
the underlying process. And in many industrial Sharing data or software code across business

Web 2019
Enabling a digital and analytics transformation in heavy-industry manufacturing
Exhibit
Exhibit 4 of 4
4

Digital assets require rigorous risk-management processes.

Digital assets require rigorous risk-management processes.
Dimensions Typical things to look at
• Clarity of the intended business (and regulatory) use of the model or use case
1 Model
context • Clarity and completeness of the initial business & technical requirements

2 Input • Appropriateness of the (development) data set

• Data quality (eg digit & duplication check) and integrity (eg missing or static data)
• Soundness of the data treatments & assumptions

3 Development
process
• Soundness and reasoning of the:
- Methodological approaches selected
- Modeling techniques used

4 Output • Quality of the model outcomes in terms of accuracy (eg back-testing &
benchmarking), precision (eg sensitivity tests) and robustness (ie model’s ability to
maintain its performance)

5 Implementation • Completeness of the model or use-case documentation

• Adequacy of the selected production environment
• Accuracy of the model code and data-import process

6 Ongoing
monitoring
• Completeness and appropriateness of the ongoing monitoring plan
• Soundness of defined KPIs and acceptance criteria for model performance

• Completeness and clarity of the model-performance reports (including consistency

7 Reporting
& use of results with model’s intended use, information on model’s performance, limitations,
weaknesses)

8 Model
governance
• Model governance (eg ownership, escalation process)
• Model controls

The model-review process defines:

• The types of review (e.g. model validation) that are expected to be performed
• Which areas of the models must be reviewed
• Who performs the review
• What expected output will result from the review
• How deep the review must be based on the materiality of the model (proportionality principle)
Depending on the implementation platform, the model review can be performed by the center of excellence
Model reviews should not be performed by the model developer

6 Enabling a heavy-industry digital and analytics transformation

units and functions can be a sensitive issue. The tools chosen should reflect the
Managers of manufacturing units can be wary organization’s wider working environment.
about revealing the true performance of their If the enterprise uses general business
operations, for example, and business leaders may applications, such as spreadsheets and
fear the exposure of commercially confidential messaging services, from one particular
information. Ensuring appropriate cybersecurity vendor, new digital applications are usually
and access-control processes from the outset of a best built on the same platforms. Indeed, the
digital transformation can allay these fears. gap between enterprise IT and OT is narrowing
in many industrial environments. Companies
Technology can use wi-fi–networked sensors for the
collection on noncritical data, for example, or
Encouraged by the vigorous marketing efforts of access process and performance data using
vendors and start-ups, it is common for companies mobile phones and tablet computers.
to take a technology-first perspective when they
think about digital transformation. They should Standardization shouldn’t keep companies
avoid that mistake. In practice, any of a number of from improving the technologies they use.
digital tools, technologies, or analytical platforms The digital space is fast-moving, with new
can be applied to a given problem. Success solutions emerging all the time. Since frontline
depends more on the organization’s ability to teams don’t have the time to keep scanning
define the problem clearly, access the appropriate the market, it’s useful to establish a corporate
data, and integrate the solution into their wider group with responsibility for identifying and
operations. evaluating promising technologies. Or to build
an ecosystem of collaborative relationships
New digital approaches don’t always require with a select group of technology providers.
significant technology upgrades. Existing
controllers and control strategies at manufacturing
sites can often be tuned or reconfigured to deliver Agile project delivery
additional value, using insights from machine- Digital projects are different from other
learning systems or related techniques. engineering efforts in heavy industry. Most
digital organizations have abandoned the
More significant than the capabilities of individual traditional waterfall engineering methodology,
digital tools is an organization’s ability to with its emphasis on formal specifications and
support their use. This is becoming especially rigid stage reviews. Instead, they adopt the
important with the increasing use of open-source agile methodology, in which development work
libraries. Companies must be very careful when takes place iteratively in short sprints, with the
managing versioning, as the open-source world is emphasis on early and continuous real-world
continuously releasing new updates that may make testing and refinement.
existing functions obsolete.
At first sight, this fast-and-loose approach to
A proliferation of approaches increases the project delivery can be challenging for teams
requirements for user training, maintenance, used to a slower, more cautious cadence.
and technical backup. It can also make it more A second look can be more reassuring. In
difficult to share successful approaches across practice, agile delivery has much in common
the business. As a result, companies should with the incremental continuous-improvement
try to standardize wherever possible, selecting cycle followed by every company that has
appropriate technologies in domains such as data taken lean manufacturing to heart. And that
analytics or machine learning after evaluating includes many heavy-industry players.
candidates across a range of use cases.

Enabling a heavy-industry digital and analytics transformation 7

 But even companies with a strong lean pedigree
will need to adopt new organizational structures
and new ways of working to accommodate agile Digitization at scale requires the right
methods. They will need to convene cross- infrastructure. The good news for heavy-industry
functional teams that include representatives players is that they already have many of the basics
from operations and IT, working alongside data in place, including a technically capable workforce,
scientists or other digital specialists. And they instrumented equipment, and a culture of
will need to adapt to the agile rhythm of quick continuous improvement. From these foundations,
development sprints, frequent prototyping, and companies can secure the talent, data, technology,
continual testing and refinement. Accordingly, and agile project-delivery capabilities they need to
large-scale transformations usually benefit from support their digital ambitions.
a dedicated project-management office to track
the progress and success of individual initiatives,
and ensure emerging best-practices are shared
across the organization.

Milan Korbel is a senior expert in McKinsey’s Melbourne office, Stuart Sim is a partner in the New Jersey office, Ken Somers
is a partner in the Antwerp office, and Joris van Niel is a senior expert in the Amsterdam office.

Copyright © 2019 McKinsey & Company. All rights reserved.

8 Enabling a heavy-industry digital and analytics transformation