For process manufacturers, analytics-infused kaizens deliver impressive results in brief time frames

A rapid improvement event, also known as kaizen, is a Lean manufacturing technique known for bringing about quick results. The traditional method, however, falls short in cases of limited visibility into production―a challenge particularly for process manufacturers monitoring multiple enclosed tanks, pipes and ongoing flows at a fast pace. Analytics-infused kaizens shed light on problems previously difficult to detect. Kaizen leaders can request precise, statistically relevant data, enabling them to go beyond hunches to pinpoint root causes and design powerful solutions aligned with high performance.

Lean manufacturing has gained renewed interest due to slower-than-expected rates of economic recovery, along with high levels of debt levels on multiple continents. These macroeconomic factors, coupled with shareholder demands for cost reduction and revenue growth, intensify the need for operational excellence and continuous improvement.

Six Sigma, which is well known for reducing variability as well as defect rates, is a proven methodology. But what if a manufacturer lacks the time and Six Sigma capabilities, particularly at a remote plant, to complete a four- to six-month Six Sigma project? In a global economy, the pressure to produce fast gains is why many manufacturers prefer to start with kaizen, long associated with generating incremental improvements in concise time frames.

Kaizen events are effective in improving discrete manufacturing workflows, as well as streamlining transactional processes. Teams typically tackle problems with limited scope and suggest improvement ideas within a week or two.

The traditional process, however, can raise more questions than provide answers for process manufacturers, whether in chemicals, energy, agribusiness, or food and beverage processing. In these industries, materials are processed continuously, from raw inputs to final output, with minimal delays. Due to multiple tanks, pipes and processes, problems in the interim stages are much less visible, and undetected problems can rapidly undermine results.

What is analytics? The management practice of analytics brings together many things—including data, statistical and quantitative analysis, explanatory and predictive models, and fact-based management—to improve the quality of business decisions. Analytics may be used as input for human decisions or for fully automated actions.1 Analytics capabilities are evolving from descriptive (what happened?) to predictive (what will happen next?). Analytics-infused kaizens provides quantitative detail that deepens insights about root causes, along with suggesting effective ways to improve process flows.

How kaizens are organizedThe traditional kaizen proceeds in three phases: preparation, kaizen and post-event. To lay the groundwork, project leaders (usually two) typically take from three to five days to prepare, depending on issue complexity, data availability and analysis.

Preparation begins with drafting a project charter, identifying primary metrics and creating a value stream map. Kaizen leaders call in subject-matter experts as needed to complete the value stream map, which can be thought of a process map with the addition of high-level data listing cycle times, changeover times, error rates or other pertinent data. Based on review of the current-state process, project leaders do an initial brainstorm and suggest hypotheses for future-state improvement.

The kaizen event, typically lasting three to five days, calls for participation from a broader team of roughly eight to 10 people, including operators running the process, as well as individuals involved in upstream and downstream activities.

The final phase is post-event, when team members take a day or two to finalize solutions. Solutions are implemented as soon as possible, or within 30-60 days, which can prove to be a huge benefit for manufacturers.

1. Davenport, Thomas H., and Jeanne G. Harris. Competing on Analytics: The New Science of Winning, Harvard Business Press, 2006.

Page 1

What an analytics-infused kaizen adds to the processKaizen events start by asking foundational questions, such as, “What is the baseline performance of our process?”and “Where are there significant delays between steps?” Analytics-infused kaizens go beyond largely qualitative and high-level analysis to add quantitative detail. Additional questions are asked:

• Isourprocessstableandpredictable?

• Aredatanormallydistributed?Ifnot,what is the optimal distribution for the data set?

• Howisourprocessperformingagainstinternal or external specifications?

• Whatdoesthedatalooklikestatisticallyin terms of mean, standard deviation and distribution?

• Dorelationshipsexistbetweenourprimary metric and one or more predictor variables? How strong are the relationships?

• Whatisthedegreeoflinearrelationshipbetween variables?

• Whichissuesoccurwithgreatestfrequency?

An analytics-infused kaizen event typically does not take longer than the traditional kaizen because it draws from data that already exist. Manufacturers with automated data collection and storage typically collect thousands of data points. It is just a question of asking for the relevant data associated with a given hypothesis.

Statistical software and an analytics platform are needed, along with basic statistical training for project leaders and in-house experts. At the very least, the team requires access to individuals with statistical skills.

Analysis of data leads either to support or rejection of a hypothesis, helping teams identify root causes. Analytics are useful both in the week of preparation and during the kaizen event, when members of the broader team may call for additional data to test additional hypotheses.

Challenging long-held assumptions In process manufacturing environments, control-room technicians and other in-house experts troubleshoot and operate their respective processes based on how they were taught, in many cases, years ago. Process control can become a routine.

One of the keys to success with kaizen is having participants with open minds and the willingness to make fact-based decisions. Being able to detect relationships between inputs, processes and output—which analytics enables—results in “aha!” moments and better prioritization of improvement ideas. Examining objective data is also a good way of diffusing conflicts among team members holding opposing viewpoints. Fact-based data analysis has the potential to break through long-held assumptions and myths.

For example, a leading ethanol producer sought to reduce high acidity levels at one of its plants. Operators assumed two factors (purge flow and Degas ratio) were responsible for high acidity. An analytics-infused kaizen event produced consensus that the solution was more likely a result of inadequately separating gases from liquid in the dehydrated alcohol condenser. The team identified piping changes to lessen acidity, and the company was also able to reduce consumption of energy by about 4 percent, resulting in thousands of dollars in savings annually.

The value of analytics-infused kaizen events can also be seen in the following two case studies.

Page 2

Case Study 1: Cornmeal producer trims enzyme use by 20 percentA producer of high-protein cornmeal used a special enzyme in production. Since plant startup, enzyme overuse resulted in higher processing costs. Operators were under the impression that any time an upset in the upstream process occurred, increasing enzyme was needed to maintain optimal protein levels.

Figure 1: Time series plot of enzyme ratio vs. process upset

80

Time Series Plot of Enzyme Ratio vs. Process Upset

Data

75

70

65

60

55

8/21

/10

9:00

AM

Empyreal Protein

Process Upset

Eight hours later showing no dramatic shift in protein level

GFD2 Protein

8/20

/10

2:00

PM

8/20

/10

3:00

PM

8/20

/10

4:00

PM

8/20

/10

5:00

PM

8/20

/10

6:00

PM

8/20

/10

7:00

PM

8/20

/10

8:00

PM

8/20

/10

9:00

PM

8/20

/10

10:0

0 PM

8/20

/10

11:0

0 PM

8/21

/10

12:0

0 AM

8/21

/10

1:00

AM

8/21

/10

2:00

AM

8/21

/10

3:00

AM

8/21

/10

4:00

AM

8/21

/10

5:00

AM

8/21

/10

6:00

AM

8/21

/10

7:00

AM

8/21

/10

8:00

AM

Accenture helped the producer organize a rapid improvement event to tackle the issue. The team began with traditional kaizen techniques—brainstorming, fishbone diagram and multi-voting—and added analytics tools.

Due to a lag in the process, process upsets from upstream would not show up in the final protein levels for approximately eight hours. At the lowest enzyme ratio used per historical

data (Figure 1), analytics showed no significant effect on final protein levels, undermining the rationale to increase enzyme dosing. Even at the lowest levels, the data revealed an opportunity for enzyme reduction without significant loss of protein. Lacking the data, it would have been difficult to persuade technicians that increasing enzyme was not needed because it was a commonly held assumption.

Page 3

Figure 2. Analysis compares reactor temperature to manufacturer’s recommendation

70

<=175 (175,178)

(178, 181)

(181, 184)

(184, 187)

(187, 190)

(190, 193)

(193, 196)

> 196

60

50

40

30

20

10

0

Process Capability for Reactor Temperature

USLOverall CapabilityPp *PPL *PPU -2.45Ppk -2.45

Process DataLSL *Target *USL 185Sample Mean 192.050Sample N 113Location 193.806Scale 1.49433

Exp. Overall Performance96 < LSL *96 > USL 99.7296 Total 99.72

Observed Performance96 < LSL *96 > USL 99.1296 Total 99.12

Expected Overall Performance is greater than 99% of the time above the manufacturer’s suggested temperature.

In addition to the enzyme analysis, operating parameters were compared to manufacturer recommendations. Process capability analysis showed the percentage of time the prime reactor temperature rose above the recommended specification (Figure 2). Analytics showed the company was operating the process above the upper specification limit more than 99 percent of the time. The

company reduced enzyme consumption by 20 percent, and implementing multiple improvement ideas resulted in annual savings of US$200,000.

Page 4

Case Study 2: Analytics-infused kaizen saves grain producer $2 millionOne of the leading US grain producers was losing hundreds of thousands of dollars annually due to lost protein. A valuable product, protein requires efficient operations to capture the maximum amount. Protein loss had been significantly higher than company standards in the prior year but in recent months it had further deteriorated.

Processes to capture and retain protein are complex with numerous unit operations, tanks, valves and piping. Minimal reliance on data and inexperienced operators led to myths, or assumptions not based on hard data, as to what was causing the loss.

Accenture helped the company conduct a kaizen event to identify root causes and suggest solutions. Based on

initial current-state information and operator experience, the team outlined relationships that might exist between input and process variables and output. The team then collected data and conducted statistical analyses.

Numerous statistical tools—including regression analysis, 2-sample t tests, analysis of variance and process capability analysis—provided insight into actual causes. Several assumptions put forth by operators were not supported by regression analysis and, therefore, were eliminated as potential causes.

Process capability analysis determined the percentage of time tank levels exceeded the overflow level. A critical tank was found to exceed the limit more than 50 percent of the time. The tank also overflowed when the transmitter read 75 percent, a key reason why operators had not detected the problem. Thermal imaging validated the finding.

Using time series analysis, Accenture showed two of three pieces of equipment began performing unfavorably at the same time the overall baseline metric began to move in an unfavorable direction. Performance on A remained relatively constant, while B and C corresponded to overall performance (Figure 3).

The team conducted additional statistical tests on the three pieces of equipment. To determine the difference in the mean performance of each, the team reviewed analysis of variance. Since the data exhibited a non-normal distribution and contained outliers, the team also conducted non-parametric testing.

Rigorous data analysis identified several causes, including which pieces of equipment were underperforming and causing protein loss. The company implemented a solution, realizing savings estimated at nearly US$2 million annually.

Figure 3. Time series chart shows trends of three critical pieces of equipment compared to an overall baseline metric.

Time Series Plot of Average Millwater, Average Clarifie, Average Clarifie, ...

Data

Month

Average Millater Spinner0.6

0.5

0.4

0.3

0.2

0.1

0

Average Clarifier A

Average Clarifier B

Average Clarifier C

Nov '10 Dec '10 Jan '11 Feb '11 Mar '11 Apr '11 May '11 Jun '11 Jul '11 Aug '11 Sep '11 Oct '11

C

B

A

Overall Performance

Page 5

Advanced analytics for high performanceContinuing to rely on long-held assumptions and intuition to solve problems is analogous to piloting a ship in a storm without modern instrumentation. Such an approach leaves root-cause analysis almost to chance, wasting time and resources.

Analytics-infused kaizen events address the limitations of the traditional method, thereby driving for clearer insight and faster results. Analytics offers improved visibility to data and reporting, along with insights for better decisions that deliver improved outcomes to the bottom line.2

While this paper has focused on analytics for process manufacturers hindered by poor visibility into production stages, analytics-infused kaizens can be used to help companies in a wide range of industries use hard data to support improvement hypotheses and produce tangible gains. Analytics can be infused into Six Sigma and other process improvement initiatives to bolster the quantitative detail that supports more precisely targeted strategies to improve speed and quality.

Organizations benefit by becoming more analytical over time. Combining advanced analytics with operators’ knowledge and process expertise improves the probability of implementing precisely targeted solutions. Advanced analytics is a leading-edge service that enables organizations to improve operations and pursue high performance.

2. Using Analytics to Achieve High Performance, Accenture 2010.

Page 6

About AccentureAccenture is a global management consulting, technology services and outsourcing company, with more than 246,000 people serving clients in more than 120 countries. Combining unparalleled experience, comprehensive capabilities across all industries and business functions, and extensive research on the world’s most successful companies, Accenture collaborates with clients to help them become high-performance businesses and governments. The company generated net revenues of US $25.5 billion for the fiscal year ended Aug. 31, 2011.

For more information, please contact:Frode Gjendem frode.huse.gjendem@accenture.com

Copyright © 2012 Accenture All rights reserved.

Accenture, its logo, and High Performance Delivered are trademarks of Accenture.