big data and big analytics for product and process quality · gather data to understand what was...
TRANSCRIPT
BigData
anD Big analytics
for Product and Process Quality
How an enterprise-wide quality platform can turn existing data into substantial and sustainable revenue growth
and cost savings for global manufacturers.
BigData
anD Big analytics
for Product and Process Quality
S2 www.sas.com I www.Industryweek.com
About IW Custom Research
IW Custom Research is an
operating unit of IndustryWeek
magazine that provides insight
into executives’ opinions
and manufacturing trends.
IndustryWeek connects decision-
makers within manufacturing
enterprises to share ideas and
tools that inspire action. In
print, online and in person,
the IndustryWeek community
is the leading resource for
manufacturing operations
knowledge. IndustryWeek is a
property of Penton Media Inc.
For more information, go to
www.industryweek.com.
About SAS®
SAS® is the leader in business
analytics software and services,
and the largest independent vendor
in the business intelligence market.
SAS’s market-leading business
analytics software and services
help customers make fact-based
decisions to improve performance,
from identifying the right product
to market to forecasting trends.
For more information, go to
www.sas.com.
About This Report
This is a report on the findings of the IW/SAS Enterprise Quality Survey*.
The objectives of the survey were to determine concerns about quality among
manufacturers; investigate the tools used to measure quality; and examine how
using enterprise-wide analysis on quality data could improve performance.
E-mail survey: On July 7, 2011, IndustryWeek e-mailed invitations to participate in
an online survey to a total 53,391 subscribers; 422 completed
surveys were returned, for an effective response rate
of .8 percent.
Title, job function, and industry: Seventy-four
percent of respondents hold executive-level or
manager positions while the rest are individual
contributors with no direct reports. Respondents
mostly work in general management (22 percent)
or operations (20 percent); 13 percent work in a
quality function; 12 percent in production; 8 percent in
supply-chain management; and the remaining in finance, IT
or “other” functions. The top industries represented are industrial machinery (21
percent), automotive (11 percent), consumer durables (9 percent), metals/mining
(8 percent) and aerospace/defense, consumer/industrial electronics, and plastics
and rubbers (7 percent each). Twenty-nine percent of represented companies
have annual revenues of $1 billion to $10 billion. The rest have annual revenues of
less than $100 million to $999 million.
*Percentages for each question do not always add up to 100 percent due to rounding or multiple responses allowed.
Contents
➤ About This Report 2
➤ Quality’s Untapped Potential 3
➤ Better Quality Through Better Data 5
➤ What is an Enterprise Analytical Quality Platform (EAQP)? 8
➤ Steelmaker Achieves $1.2 Million ROI with EAQP 8
➤ Key Benefits: Time-to-Market, Problem Resolution, Performance Improvement 9
➤ Conclusion 11
www.sas.com I www.Industryweek.com S3
Quality’s untapped Potential
Quality is not a difficult concept
for manufacturers or their
customers to understand. If the
customer is 100 percent satisfied, the
manufacturer has achieved the highest
level of quality. Of course, this isn’t
possible all of the time. Even Six Sigma
allows room for some error. Quality,
then, is something that manufacturers
constantly work at, and the most
commonly used tool for this work is data.
Hundreds of everyday manufacturing
metrics measure some aspect of quality:
➤What’s this week’s scrap
rate compared with last week’s?
➤Have our warranty costs changed
since we switched suppliers?
➤ If we shorten order-to-delivery cycle
time, will we increase market share?
Customers don’t see all of the
measuring, inputting and analysis that
goes into quality assessment — which
is good. But for a variety of reasons,
access to and analysis of quality-related
data tends to be limited internally as
well — and that’s not good. This research
report explains why and describes
how to turn this common shortcoming
into a competitive advantage with an
enterprise-wide analytical solution.
IndustryWeek and SAS partnered to
delve deeper into how manufacturers
are using and not using quality-related
data to improve performance because
we’ve recognized that companies need
to change their thinking on quality in
order to address the ongoing challenges
of competing in a global economy.
In the past, quality improvements
occurred at a very granular level.
Teams would uncover defects and then
gather data to understand what was
going on with either the product design
or the production process. At the base
level, companies got serious about
quality when they introduced statistical
process control, which enabled them to
have some warning of processes moving
out of control. This was the first step.
Acknowledging the interdependence of
variables that contribute to yield and
quality objectives, manufacturers
took the next step in developing
their information systems: process
engineering. Process engineering
recognizes relationships among multiple
inputs and makes progress toward
assembling and displaying data in
cross-functional context. It disseminates
information faster to those who will take
corrective or optimization actions.
But might there be a third step?
Could we move from process
engineering to process
understanding? A holistic view
across the entire organization? A
high-level view distilled from the
lowest-level details? What if a
company could have strategic,
predictive and optimal vision based
on one version of the truth, assembled
from all relevant process, product,
equipment and service data — an
information framework that recognizes
the fundamental interdependence of
functions and metrics that were once
viewed in isolation?
“The individual quality
project is still very much
a part of the process.
Global collaboration does
not preclude innovative
process improvement
through an individual
engineer’s efforts. But with
an enterprise platform,
engineers can share best
practices, successful
models and techniques,
as well as data, thereby
increasing the value of
quality engineering to the
company.” — michael newkirk, manufacturing and supply chain marketing manager for sas
S4 www.sas.com I www.Industryweek.com
This is possible when an organization takes an
enterprise-wide or global approach to collecting
and analyzing quality-related data, and begins
to see this data not only as a quality-control tool,
but also as a strategy to guide action in key areas
that feed fast and direct results to the top and
bottom lines.
“Culturally, looking at process improvement
on a global basis is a tough sell,” said Michael
Newkirk, Manufacturing and Supply Chain
Marketing Manager for SAS. “So much process
improvement has occurred by decentralizing
and de-bureaucratizing quality initiatives.
Leaders and managers think, ‘By the time
we got all that data together, our individual
engineers could find and solve dozens of profit-
draining problems.’ That might have been
the case a decade ago, but global process
improvement is very attainable today. ”
As the IW/SAS Enterprise Quality Survey and
this report show, enterprise-wide quality is an
untapped gold mine for the modern, global
manufacturing company. Using an Enterprise
Analytical Quality Platform (EAQP), companies
can unify the quality-related data that they
already collect and analyze it in multiple
ways to identify new opportunities to reduce
costs and improve profits; and to prioritize
improvement opportunities by determining
which will result in the fastest and greatest ROI.
“The most sophisticated manufacturers are
already doing this,” Newkirk said. “They have
created a living archive of collective knowledge,
where data from across the organization
is warehoused together and accessible to
everyone working on process and product
improvement. The results of individual efforts
are published and scored. Models that work
the best are distributed while those that don’t
are discarded.
“There is, in essence, a virtual workbench
created for collaboration across the entire
company. Best practices, best models, design-
of-experiment results — all can be archived
and shared.”
Most importantly, according to Newkirk, the
individual quality project is still very much a part
of the process. The enterprise platform, while
being extendable and scalable — and capable
of handling deep analysis on large amounts of
data, such as data mining, text mining and high-
performance forecasting — also gives individual
engineers and analysts desktop copies of
analytical tools and visualization capabilities.
With these they can design experiments, and
visualize data, scenarios and outcomes; and
then share their results and models with the rest
of the enterprise.
“Global collaboration does not preclude
innovative process improvement through
an individual engineer’s efforts. But with an
enterprise platform, engineers can share best
practices, successful models and techniques,
as well as data, thereby increasing the value of
quality engineering to the company. ”
In this report, we will explain how an EAQP
works and how using such a platform can
accelerate time-to-market for new products;
reduce the negative effects of existing problems
by resolving them faster; and enable long-term
sustainment of quality improvements. We will
also tie these benefits to the results of the
IW/SAS Enterprise Quality Survey, which reveal
a need for such a platform in all sectors and
sizes of manufacturing enterprises.
As the IW/SAS
Enterprise Quality
Survey and
this report show,
enterprise-wide
quality is an
untapped gold mine
for the modern,
global manufacturing
company.
www.sas.com I www.Industryweek.com S5
Manufacturers put an enormous amount of work into improving quality. Yet, for all of the
effort, they still struggle. Thirty-eight percent of respondents to the IW/SAS Enterprise
Quality Survey have seen the cost of quality as a percentage of cost-of-goods-sold increase
over the past two years; and it stayed the same for 28 percent of respondents. (See Figure 1.)
Better Quality through Better data
0% 10% 20% 30% 40% 50%
Change in COGSHow has your company’s quality costs as a percentage of COGS changed over the past two years?
Increased more than 20%
Increased 11%-20%
Increased 6%-10%
Increased 1%-5%
Stayed the same
Decreased 1%-5%
Decreased 6%-10%
Decreased 11%-20%
Decreased more than 20%
None of these
No reply
1%
2%
11%
24%
28%
19%
5%
2%
1%
5%
1%
0% 20% 40% 60% 80% 100%
Quality ChallengesWhich of these are your company’s biggest quality challenges? (Select up to three.)
Reducing scrap and rework costs
Sustaining desirable quality and/or yield levels
Having a sufficient amount of quality- related data to provide meaningful insight
Reducing warranty-claims cost
Our quality-related data is inconsistent, outdated or otherwise flawed
Other
57%
49%
22%
20%
10%
7%
FIGuRe 1
FIGuRe 2
More than half of respondents
identified reducing scrap and
rework costs as a top quality
challenge. (See Figure 2.)
About half identified sustaining
desirable quality and yield levels
as a top challenge, as well. The
third-most-cited challenge was
having a sufficient amount of
quality-related data to provide
meaningful insight.
We believe that manufacturers
are being held back on excelling
at the first two challenges
by the third. Our research
shows that only a small slice
of manufacturing employees
review quality data. Not
surprisingly, it’s common for
operations managers and quality
engineers to review the data.
(See Figure 3, page S6.) But
only one in three continuous-
improvement managers look at
the data, and only one in four
customer-facing employees
such as salespeople do. Only
27 percent of interviewees said
everyone at their company
reviews quality data.
S6 www.sas.com I www.Industryweek.com
In general, manufacturers could do a better job
of data integration, analysis and accessibility,
a fact that is particularly true for quality-related
data that could provide strategy-level insight.
This is because vast amounts of quality-related
data are hidden by the increasing complexity
of IT assets and the confusion and frustration
that results. It’s common at most companies —
regardless of size — for leaders and non-IT staff
members to assume that the data reports they
are given by the IT staff are all they need to work
effectively and all that is available. Not true. But
to change this environment, company leaders
will need to see the true strategic value in the
data they already collect.
“[Leaders] simply outsource the work of strategic
technology design to contractors and IT staff and
hope that the chaos gets sorted out over the years.
As a leader, you cannot afford to do this anymore.
You must have a strong working knowledge of
your digital plumbing before you can begin to
optimize its performance, and the performance of
your organization.” Scott Klososky, The
Velocity Manifesto1
Looking more closely at the IT landscape in a
manufacturing environment, it’s not uncommon
to find a multitude of different systems from
which data is extracted:
➤ERP/MRP systems: Often there are
multiple installations from one vendor, but all
are on different release levels, and they have
differing customizations. Because of mergers
and acquisitions, it is common to find
multiple vendors compounding the problem
of data access.
➤MES: A company with multiple
manufacturing facilities most likely has a
range of different Manufacturing Execution
System data sources. Harmonizing the MES
to a corporate standard is typically too huge
of an effort and is likely to cause production
disruptions that cannot be afforded. Also,
MES implementations have the same problem
that ERP/MRP implementations have with
0% 20% 40% 60% 80% 100%
Review of Quality-Related InformationWho reviews quality-related information (i.e., scrap, rework, warranty claims)at your company/plant?
Operations managers and supervisors
Quality engineers
Continuous-improvement leaders
Product engineers
everyone
Customer-facing functions such as sales
Procurement/purchasing professionals
Plant and maintenance engineers
Other
None of these
71%
52%
33%
30%
27%
24%
23%
14%
5%
1%
FIGuRe 3
1 Klososky, Scott, The Velocity Manifesto, Greenleaf Book Group Press, Austin, Texas, 2011
www.sas.com I www.Industryweek.com S7
customizations, even using standardized
platforms. That is, even using the same
platform from plant to plant, the data can
be very hard to extract.
➤Testing: Test systems typically come
with a proprietary data format and storage
applications. Having acquired testers from
multiple vendors for valid purposes adds to
this complexity. Test data often are collected
into Quality Management Systems (QMS)
and Laboratory Information Management
Systems (LIMS) that are plant-specific and
difficult to share system wide.
➤Equipment: With more sensors and
microprocessors being engineered into
production assets, process equipment
and machinery is collecting an increasing
amount of data that could be used
to improve equipment performance,
processes and yields. This data is under-
utilized. Plant and facility engineers usually
use the data only for historical evaluations,
although it could be used for predictive
analysis.
➤Defects: Manufacturers perform
inspections on incoming material, the
semi-finished and the finished products,
and they use some sort of tracking and
cataloging of symptoms, defects and actions,
such as equipment repairs, adjustments to
processes or product redesigns. This is key
quality data, and having access to this and
similar data and analyses across other lines
and factories could have enormous benefits.
➤Process: Data from manufacturing
operations is collected from a variety
of shop-floor data sources such as
Programmable Logic Controllers (PLC),
RFID tags, bar code scanners and manual
data input stations (HMI/MMI), but it is not
aggregated and assembled in a way that
facilitates quality exploration and analysis.
➤Supplier: When manufacturers
receive raw material, components or sub-
assemblies from suppliers, they often
receive quality data along with products.
This information gets stored and is valuable
for analysis but not always easy for quality
analysts and engineers to access.
In such a complex environment, the process
of getting at the data is very tedious and
time consuming, and probably not very
repeatable. This is true for repetitive and
necessary tasks, such as root-cause analysis
and standard reporting. And for strategic
insight that requires real-time to near-real-
time data, gathering enough data in a timely
enough fashion for it to be meaningful and
actionable is impossible.
Yet, such a need is obvious among
respondents to our survey. When we
asked participants what would improve
their company/plant’s quality performance
as a contributor to profitability, here is how
they responded:
➤Early-warning analytics that enable
operators to proactively address potential
quality and performance issues before they
become customer problems: 49 percent.
➤Collaborative problem-solving
capabilities across departmental functions
through enterprise-wide technology tools:
49 percent.
➤Easier and faster access to quality-
related data and other knowledge:
47 percent.
➤Predictive modeling that allows optimal
process set-up, leading to improved asset
utilization, optimized material consumption,
reduced rework rates, and reduced scrap
expenses: 37 percent.
An EAQP can provide all of these
capabilities and elevate the quality data you
already collect from granular and tactical to
insightful and strategic.
In such a complex
environment, the process
of getting at the data is
very tedious and time
consuming, and probably
not very repeatable. This
is true for repetitive and
necessary tasks, such as
root-cause analysis and
standard reporting. And
for strategic insight that
requires real-time to
near-time data, gathering
enough data in a timely
enough fashion for it to be
meaningful and actionable
is impossible.
STeelmAkeR ACHIeveS $1.2 mIllION ROI wITH eAQP
A korean steelmaker had made sound incremental process improvement using a cadre of highly trained six
sigma black belts. But they sensed they could make a larger impact with an enter-prise-wide, coordinated effort that utilized data integration and enterprise data mining and modeling.
there were still large profit variables among plants and items; and scrap losses were unacceptably high. traditional analysis on isolated processes wasn’t sufficient.
By pulling all of its data together across plants and processes, the company re-duced its scrap ratio from 15 percent to 1.5 percent, saving $150,000 on one process. the company was able to identify variations in profitability by plant and item for cold roll steel, delivering an annual $1.2 million return on its investment in the soft-ware solution that the six sigma experts used to identify high-value improvement targets that teams then analyzed for root-cause problems.
the steelmaker also achieved a 50 per-cent reduction in lead times for standard hot coil production (from 30 to 14 days) and reduced inventory by 60 percent. these efforts helped reduce the planning and sales cycles, and maximized production utilization. the improvements also didn’t take months to accomplish — the analyti-cal cycle times actually dropped. 2
S8 www.sas.com I www.Industryweek.com
Figure 4 shows a conceptual
architecture of an enterprise
solution. It shows:
The Information Store: The data
sources on the left are the most
important building blocks for the EAQP
data model. The rest of the data sources
are augmenting the value of the solution.
Quality Lifecycle Analysis: This layer
includes multiple interfaces suited for
different types of users.
The interfaces could include:
➤ Dashboard to surface KPIs.
➤ Reports to fulfill an organization’s
reporting requirements.
FIGuRe 4
2 “How This Giant is Light on Its Feet: Korean steel giant POSCO achieves Six Sigma performance with the help of SAS®,” SAS Institute Inc., www.sas.com
what is an enterprise analytical Quality Platform (eaQP)?
Enterprise Analytical Platform Overview
Parts movement
measurements
Testing
Inspection
Products & Hierarchies
Finance & Organizational
equipment
Repair & FA
Field Failures
Supplier Data
➤ Alerts to surface continuous-monitoring results.
➤ Analytical workbench for root-cause analysis.
➤ Predictive modeling capabilities to allow proactive quality management.
➤ Model management to support the processes around predictive
models, including performance monitoring.
Users gain these benefits:
➤ Access to and integration of key data from multiple sources, platforms, and operational systems.
➤ Optimal analytical use of data.
➤ Creation of a single version of the truth for the entire organization.
➤ Disbursement of best-practices-based business rules and processes. These are applied to all
reports and analyses.
www.sas.com I www.Industryweek.com S9
While the benefits of using an EAQP are many, the platform’s potential in three areas is key to manufacturing competitiveness: time-to-market, problem solv-ing and performance improvement. Let’s take a look at each of these. In the
following charts, the curves depict the lifecycle of a product.
Finance & Organizational
equipment
Repair & FA
Field Failures
Supplier Data
key Benefits: time-to-market, Problem resolution, Performance Improvement
Figure 5 shows that during development, a company invests heavily to develop a new product/model, and the early proto-types typically show low yields and are not sellable product. Once the company completes development and is ready to bring the new product/model to the market, it must introduce new processes into production.
Initially yields tend to be quite low at this point, and the goal is to bring them up to optimal levels. At the same time, production volumes need to be in-creased. In many companies, the ramp up of volumes and yields is moderate.
During the early part of a product’s lifecycle, the market will tolerate a higher price, but as competition quickly catches up, the street price for an item declines. This means that the time span to recover the upfront R&D investment is quite limited. The faster a company can ramp up volumes and yields, the more recovery of R&D investment and profit can be made.
An EAQP enables companies to signifi-cantly reduce the time-to-mature-yields during the yield-ramp up phase through: ➤ Swift feedback. ➤ Rich analytical toolset. ➤ Increased transparency and insight into manufacturing processes.
Once products have reached high volumes and mature yields, a variety of unexpected occurrences can cause a yield or quality problem to creep up, as is depicted in Figure 6. In these cases, it’s essential to detect the issue as fast as possible, uncover the root cause(s) and resolve the problem.
FIGuRe 5Time-to-Market
YIE
LD
TIME
DevelopmentYield Ramp
Volume Production
FIGuRe 6Problem Resolution
YIE
LD
TIME
DevelopmentYield Ramp
Volume Production
S10 www.sas.com I www.Industryweek.com
This is because during a period of low yields/bad quality, a producer could be: ➤ Losing production capacity. ➤ Missing shipments. ➤ Incurring excessive costs. ➤ Exposing customers to faulty products. ➤ Facing future financial liabilities because of warranties, recalls and/or loss of goodwill.
An EAQP allows you to significantly reduce the amount of time it takes to resolve problems through monitoring and early-warning capabilities, and ana-lytical root-cause analysis capabilities. This means that high levels of quality can more easily be sustained. Sustain-ing desirable quality and/or yield levels was the second-largest quality chal-lenge survey respondents named at 49%. (See Figure 2, Page 5.)
Once yields have reached mature lev-els, companies might want to push the limits to increase output and produc-tivity — without needing to invest in additional production capacity. (See Figure 7.) This is hard work, and to achieve successful and sustain-able results, an organization needs advanced statistical discovery capa-bilities to uncover where the greatest improvement opportunities lie.
FIGuRe 7Performance Improvement
YIE
LD
TIME
DevelopmentYield Ramp
Volume Production
*A decrease in time-to-market is an improvement; an increase is a decline in quality.
Companies with higher levels of data integration and sharing had greater improvement in quality, customer satisfaction and time-to-market over the past two years.
Data integration of/accessibility to product-related data ➔
High Some None Data sharing between upstream/ downstream processes ➔
High Some None
How has this metric changed in past 2 years? (estimated mean shown)Quality as % of COGS -0.8% +0.6% +1.6% -0.1% +0.8% +3.6%Customer satisfaction +4.9% +3.9% +3.1% +5.1% +3.3% +0.8%Time-to-market for new products*
-3.7% -3% -2.6% -4.7% -1.9% -0.5%
FIGuRe 8
An EAQP allows engineers and analysts to find the opportunities that will return the greatest gains for the effort by using hard statistical facts through a rich port-folio of analytical discovery tools. Because they have the data available and can collaborate on a common platform, these opportunities come fast enough to be realistically actionable. For all the investments of time, labor and sometimes equipment that perfor-mance improvement requires, choos-ing where to make the improvements
should not be left to “gut feel.”
Findings from the IW/SAS Enterprise Quality Survey also make a connection between these three areas and having highly integrated, widely accessible enterprise data. In Figure 8, notice that respondents whose companies have highly accessible, highly integrated data across the enterprise have reduced time-to-market, decreased COGS, and improved customer satisfaction more during the past two years.
www.sas.com I www.Industryweek.com S11
Despite rapid advances in technology in all areas of business, companies have focused on communications tools the most and have greatly underestimated data analysis as a competitive weapon. 3
Quality data, which abounds in manufacturing enterprises, is one of the largest caches of untapped potential in this area and can be elevated to a strategic enabler by imple-menting an Enterprise Analytical Quality Platform (EAQP).
3 “Gearing For Growth: Future drivers of corporate productivity,” Economist Intelligence Unit, wwweiu.com
*A decrease in time-to-market is an improvement; an increase is a decline in quality.
Companies with higher levels of data integration and sharing had greater improvement in quality, customer satisfaction and time-to-market over the past two years.
Data integration of/accessibility to product-related data ➔
High Some None Data sharing between upstream/ downstream processes ➔
High Some None
How has this metric changed in past 2 years? (estimated mean shown)Quality as % of COGS -0.8% +0.6% +1.6% -0.1% +0.8% +3.6%Customer satisfaction +4.9% +3.9% +3.1% +5.1% +3.3% +0.8%Time-to-market for new products*
-3.7% -3% -2.6% -4.7% -1.9% -0.5%
conclusion
As leading companies have demonstrated, an EAQP creates a globalized manu-facturing intelligence architecture that returns swift and significant benefits with actionable and sustainable results. (See Figure 9.)
Companies, particularly manufacturing companies, will face continued pressure to improve performance with existing resources in order to remain competitive. C-level executives want a greater return on the 30-plus years of investment into operational and transactional IT systems. Unlocking the potential of existing quali-ty-related data can do this while supporting ongoing process-improvement efforts.
For more information on EAQP or other quality-related solutions, contact www.support.sas.com.
FIGuRe 9Manufacturing Intelligence Architecture
S12 www.sas.com I www.Industryweek.comSAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc. in the USA and other countries. ® indicates USA registration. 105414_S80746.1011