Playing catch-up in advanced analytics
Among
CPG companies’ most underused assets are the vast quantities of data
they generate. But there’s still opportunity to invest in
industry-beating analytic capabilities.
Consumer-packaged-goods
(CPG) companies have increasingly
gained access to unprecedented amounts of data, and we expect that this trend
will intensify over time. However, most companies are generating very limited
insight from these newly found sources, leaving a treasure chest of
opportunities untapped. Companies that have already captured benefits from this
trove of insight are reaching the next frontier of performance. For those
companies that have yet to follow suit, it is imperative to take steps to catch
up.
With all the attention given to
advanced analytics, it may be tempting to
assume that the benefits are overrated. But the value is real and significant.
For example, by making a concerted effort to find correlations among more than
400 output and quality parameters, a paper company was able to reduce material
cost by more than 10 percent and increase revenue per ton by more than 5
percent. To achieve its targets for quality at minimum cost, the company needed
to optimize multiple consecutive process steps. Papermaking is a complex
industrial process that combines interrelated chemical, thermal, and mechanical
steps, but these interrelations and their impact on the product are often not
well understood. The company had only a limited understanding of how control
parameters influence product characteristics at each step of the value chain.
Once the company understood the linkages among parameters, it used the
latest machine learning and bespoke algorithms to
optimize the paper mill for customer requirements. It also uses the model on a
continuous basis to monitor and increase yield and improve performance.
Choose the right data and tools
Global companies should
consider using both internal and external data, while addressing the challenge
of accessing information in the most sensible way. Internal data sets must
typically be cleaned and merged in order to glean insights. Depending on their
source, data sets generated within a plant are often structured differently
(for example, using different nomenclature or units of measure) and may contain
statistical “noise” for a variety of reasons. External data can often provide
CPG companies with much-needed insights related to consumers, customers, and
competitors. However, the ability to use external data sources is highly
dependent on the extent to which the provider has structured the data. Raw data
may appear to be a random array of numbers and names. To make the data useful
for analysis, the provider must, for example, clearly identify what is being
measured and the units of measure. To generate insights from the data sources,
companies must also understand the analytical tools and models available and
select the right ones.
By choosing the right
reliability-analytics approach, a heavy-equipment manufacturer reduced
maintenance spending by about 5 percent, amounting to savings of tens of
millions of dollars. The company applied a variety of advanced analytics to
evaluate the performance of large motors used in production. These analytics
included a Weibull analysis of component-failures data to determine the optimal
maintenance strategy and logistic regression analysis to establish the
relationship between alarms and failure events. Although these are not new analyses,
the availability of real-time data, on-demand computing power, and
machine-learning algorithms has enabled an entirely new level of proactive
insight. The analyses revealed that high maintenance spending resulted from the
need for a break-in period for new components. To prevent breakdowns, the
manufacturer had been proactively replacing the components. In fact, the
company’s overzealous approach to component replacement had actually been
driving costs higher. By replacing the components less frequently, the company
could improve their reliability and reduce maintenance costs.
Take a comprehensive approach
Manufacturers must apply a robust
advanced-analytics approach that
includes statistical analysis, modeling, and optimization of processes and
products, and then take action based on the insights:
·
Collect data through sensors along the entire manufacturing
process at critical points.
·
Record, archive, and
analyze large volumes of big data using
statistical analyses.
·
Model, using logic and
algorithms, to calculate optimal settings or modes
of operation. The state-of-the-art approach is to combine predictive methods into
one superior method to model insights. The combination of methods, called
blending, is itself a complex algorithm searching for the best possible
solution. The predictive methods blended in this approach are often not new.
However, the availability of vast amounts of additional data and tremendous
processing power has created new opportunities for companies to combine these
analyses to predict performance and determine the optimum answers to business
problems.
·
Act on the optimization and
continually monitor the impact to ensure decisions in
plants are consistent with the design.
·
Set up and/or adjust the
process or equipment with continuous improvement.
Across CPG industries,
companies are capturing significant value by deploying an analytics approach based
on these steps. Two examples illustrate the potential:
Improving vaccine
yield. A biopharma
company increased its processing yield by 50 percent by building artificial
neural networks to identify optimal machine parameters. Manufacturing was the
bottleneck in the company’s value stream. Yield was declining and highly
variable, and regulatory bodies had raised concerns about process instability.
The company’s existing approach to improve yield was unstructured and showed
limited success. The company adopted a new approach that began with thoroughly
understanding its manufacturing process. It gathered all available historical
process data sets and used advanced analytics to identify the most influential
parameters. It applied the insights to identify linkages between critical
parameters and operations, define the main levers for reaching the optimum
process conditions, and develop an action plan for implementing the levers.
Without the need for capital expenditures, the company succeeded in
de-bottlenecking its production facility. The yield improvement helped to
reduce costs by $5 million to $10 million and increase sales by $20 million.
The improvement also led to a resolution of the regulatory issues.
Accurately predicting
spiciness. A food
manufacturer was able to raise quality and reduce complaints by 90 percent by
using infrared (IR) technology to continually monitor quality and optimize
parameters to improve flavor characteristics. The perceived “hotness” of the
company’s products is a critical parameter, but could only be determined by
human testers—a process that is both unreliable and expensive. The company
sought to use the latest IR technology to capture data, and then use data
analytics to develop a model to evaluate hotness. The company developed a fully
automated model in a neural-network tool to correlate key IR measurements to
hotness. It validated the model through production runs and a comparison with
human flavor-tester results. Once data capturing and modeling was validated,
key-input-controlling parameters were connected through a process loop,
enabling the manufacturer to change inputs to meet hotness targets. The
manufacturer now continuously monitors quality prior to packaging. Total
testing costs have been reduced, while the sampling size has increased and
variations in product characteristics have declined.
Efforts to adopt
advanced analytics are typically initiated by teams in the
industrial-engineering or continuous-improvement functions. While the value
captured is often significant, these efforts are typically isolated and
uncoordinated. Embedding such approaches more systematically requires
coordination by the COO, working with other members of the C-suite to ensure
that adequate energy, resources, and attention are brought to bear. Most
manufacturers will find that the data they need in order to derive valuable
insights is already at their fingertips. Applying the right approach to
advanced analytics is the crucial step to catching up to competitors that have
already reached the next frontier of improvement.
By Luis Benavides, Rehana Khanam, Frédéric
Lefort, and Oscar Lovera-Perez
http://www.mckinsey.com/industries/industries/consumer-packaged-goods/our-insights/playing-catch-up-in-advanced-analytics?cid=other-eml-alt-mip-mck-oth-1705&hlkid=2e8cedcff7b948b0b314a79bdd3e5dfc&hctky=1627601&hdpid=1feb516e-c5db-4344-b8f2-428eb5c38d1e
No comments:
Post a Comment