Journal Pre-proof

Identifying pathways to a high-performing lean automation implementation: An

empirical study in the manufacturing industry

Guilherme Luz Tortorella, Gopalakrishnan Narayanamurthy, Matthias Thurer

PII: S0925-5273(20)30275-9
DOI: https://doi.org/10.1016/j.ijpe.2020.107918
Reference: PROECO 107918

To appear in: International Journal of Production Economics

Received Date: 16 April 2020

Revised Date: 24 July 2020
Accepted Date: 29 August 2020

Please cite this article as: Tortorella, G.L., Narayanamurthy, G., Thurer, M., Identifying pathways to
a high-performing lean automation implementation: An empirical study in the manufacturing industry,
International Journal of Production Economics (2020), doi: https://doi.org/10.1016/j.ijpe.2020.107918.

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.

© 2020 Published by Elsevier B.V.

 Identifying pathways to a high-performing Lean Automation

implementation: an empirical study in the manufacturing industry

Guilherme Luz Tortorella* (gtortorella@bol.com.br)

The University of Melbourne, Melbourne – Australia

Universidade Federal de Santa Catarina, Florianópolis – Brazil

of
ro
Gopalakrishnan Narayanamurthy (g.narayanamurthy@liverpool.ac.uk)
-p
University of Liverpool, Liverpool – UK
re
lP

Matthias Thurer (matthiasthurer@workloadcontrol.com)

na

Jinan University, Zhuhai – China

ur
Jo

* Corresponding author

Abstract

This paper examines pathways to implement a high-performing Lean Automation (LA). We

asked 61 manufacturers from Brazil and India that are undergoing a lean implementation

together with the adoption of disruptive digital technologies from Industry 4.0 (I4.0) to

indicate their implementation sequence. We then used multivariate data techniques to analyze

the collected data. Our findings suggested three sets of lean practices and I4.0 technologies;

1
namely: start-up, in-transition and advanced. Further, companies that presented a higher

performance improvement have more extensively implemented start-up and in-transition

practices/technologies. However, no significant difference was found for the adoption level

of advanced practices/technologies between low- and high-performer companies. Since the

integration of I4.0 technologies into Lean Manufacturing (LM) is a relatively recent

phenomenon, our study provides guidelines related to a preferential implementation sequence

within this portfolio of practices and technologies.

Keywords: Lean Automation, Lean Manufacturing, Industry 4.0, Performance.

of
ro
-p
re
lP
na
ur
Jo

2
 Identifying pathways to a high-performing Lean Automation

implementation: an empirical study in the manufacturing industry

Abstract

This paper examines pathways to implement a high-performing Lean Automation (LA). We

asked 61 manufacturers from Brazil and India that are undergoing a lean implementation

together with the adoption of disruptive digital technologies from Industry 4.0 (I4.0) to

of
indicate their implementation sequence. We then used multivariate data techniques to analyze

ro
the collected data. Our findings suggested three sets of lean practices and I4.0 technologies;
-p
namely: start-up, in-transition and advanced. Further, companies that presented a higher
re
performance improvement have more extensively implemented start-up and in-transition
lP

practices/technologies. However, no significant difference was found for the adoption level
na

of advanced practices/technologies between low- and high-performer companies. Since the

ur

integration of I4.0 technologies into Lean Manufacturing (LM) is a relatively recent

Jo

phenomenon, our study provides guidelines related to a preferential implementation sequence

within this portfolio of practices and technologies.

Keywords: Lean Automation, Lean Manufacturing, Industry 4.0, Performance.

1. Introduction

Lean thinking has been subscribed for decades by both manufacturing and services sector to

improve their performance (Womack and Jones, 1997; Stone, 2012). One of the strong

reasons for the consistent adoption of lean thinking across diverse sectors is the simplicity at

which the different lean tools and techniques can be implemented even by the shop floor

1
employees, by just relying on common sense (Holt, 2019). Lean Manufacturing (LM) was not

only simple to implement but also delivered large returns for firms. It helped firms to

significantly reduce non-value adding tasks and enhance value-adding tasks, which finally

enhanced their operational performance (Shah and Ward, 2003; Chavez et al., 2013;

Bortolotti et al., 2015a).

In the recent past, firms have started adopting Industry 4.0 (I4.0) by deploying smart

components and machines that are integrated into a common digital network based on well-

proven internet standards (Kolberg et al., 2017). Researchers have stated I4.0 as a new

of
industrial paradigm that can enable firms to deliver higher financial, ecological and social

ro
performance (Stock et al., 2018). Through the deployment of digital technologies, I4.0
-p
facilitates higher levels of mass customized processes, products and services (Zawadzki and
re
Żywicki, 2016), new product and service developments (Dalenogare et al., 2018), and
lP

business model innovations (Frank et al., 2019) allowing firms to achieve improved
na

performance levels.
ur

As both LM and I4.0 have individually shown to enhance performance, firms have started
Jo

integrating both approaches to achieve superior performance and competitive advantage over

their competitors in the market (Tortorella et al., 2019a; 2019b). On the one hand, LM

delivers its positive impact on performance through a systematic and continuous search for

waste reduction and improvements (Narayanamurthy and Gurumurthy, 2016). On the other

hand, I4.0 technologies introduce automation and interconnectivity that can mitigate pre-

existing management difficulties (Tortorella et al., 2020a). Combining LM with I4.0 helps

firms in achieving Lean Automation (LA), which according to Kolberg et al. (2017) aims for

higher changeability and shorter information flows to meet future market demands.

Therefore, it is clear that these two interventions introduce capabilities that can operate

together to lead firms to new performance standards that are much higher than in the past.

2
However, even though literature converges on the potential of combining LM and I4.0

implementation together to achieve LA for higher performance, research has not focused on

examining if the sequence of implementation of different LM practices and I4.0 technologies

in a firm will have an impact on its operational performance. That the actual implementation

sequence has an impact is indicated by Browning and Heath (2009) in the context of LM

practices. Browning and Heath (2009) conducted a detailed case study research of Lockheed

Martin’s production system for the F22. They proposed that the cost reduction benefits

achievable through the implementation of LM practices could vary depending on the timing

of
of their implementation. This temporal aspect of LM practices implementation played such a

ro
crucial role that it can change the benefits to go from positive to negative (costs).
-p
Similarly, research has discussed implementation patterns of I4.0 technologies. For instance,
re
recently Frank et al. (2019) started by splitting I4.0 technologies into two broad categories,
lP

namely front-end (comprising of smart manufacturing, smart products, smart supply chain
na

and smart working) and base technologies (comprising of cloud, IoT, big data and analytics).
ur

Then by applying cluster analysis, they defined patterns of adoption of these two layers of
Jo

technologies in the surveyed companies and summarized the sequence of implementation in a

framework. Yet, while the importance of the implementation sequence is widely recognized,

no study to date empirically assessed the sequence in which LM and I4.0 are implemented in

practice, and how the different pathways affect performance. Therefore, by extending the

research on finding the optimal sequence of implementation of LM practices and I4.0

technologies individually, we answer the following research question:

RQ: Which pathway of implementation of LM practices and I4.0 technologies can

mature the LA intervention to achieve superior performance improvement?

To answer the above-stated research question, we randomly listed 21 LM practices suggested

by Shah and Ward (2003) and 14 I4.0 technologies listed by Tortorella and Fettermann

3
(2018) and Rossini et al. (2019). The respondents were asked to sequence the LM practices

and I4.0 technologies in the order of their implementation and provide the response on their

adoption level. Finally, respondents’ perceptions on operational performance improvement

during the last three years were recorded. The data was collected from manufacturing firms in

India and Brazil using a questionnaire. The final sample comprised of 61 manufacturers from

Brazil and India that were undergoing a lean implementation together with the adoption of

digital technologies from I4.0. We used multivariate data techniques to analyze the collected

data.

of
Our findings indicate that there is no clear path in terms of individual practices/technologies.

ro
Yet, there are sets of LA practices/technologies that are more prone to be implemented first
-p
than others, suggesting the existence of a precedence relationship. Our study identifies three
re
sets of lean practices and I4.0 technologies using unique ranking data. Since the integration of
lP

I4.0 technologies into lean management is a relatively recent phenomenon, our study
na

provides guidelines related to a preferential implementation sequence within this portfolio of

ur

practices and technologies. To the best of our knowledge, this study is the first one to use
Jo

ranking data to answer the aforementioned question.

The paper is structured as follows. In section 2, we review the related literature to provide the

background to our study. In section 3, we present the research method, including

questionnaire development, data collection, and data analysis. In section 4, we discuss the

results obtained and develop a schematic representation of the pathway to a high-performing

LA implementation. Finally, in section 5, we conclude the research, outline implications for

research and practice, and list the limitations and future research directions.

2. Background

4
In this section, we review literature that has discussed the importance of the sequence in

which LM and I4.0 are implemented. Following it, we then review the literature that

discusses the impact of LA on performance. Note that we do not aim to provide a systematic

and comprehensive review of the literature. This would be far out of the scope of this study

given the large amount of relevant literature. We rather seek to identify and discuss key

papers to contextualize our study.

of
2.1. LM implementation

ro
LM implementation is grounded on five broad tenets: (i) identify/define value, (ii) map the
-p
value stream, (iii) create flow, (iv) establish pull system, and (v) pursue perfection. Adopting
re
these tenets in sequence allows value stream managers to spot inefficiencies, create better
lP

flow in work processes, enhance value for their customers, and develop a continuous

improvement culture (Åhlström, 1998; Åhlström and Karlsson, 2000). It has been noticed in
na

recent research that these tenets are operationalized differently by firms leading to a
ur

disconnect between theory and practice (Narayanamurthy et al., 2018).

Jo

According to Browning and Heath (2009), maturation of lean research has advanced more

rapidly in philosophy than in actual theory and the mechanisms governing how and when to

apply lean practices require further elucidation. The majority of lean practices help in

achieving one of the two key objectives – controlling inventory buildup and reducing system

variability. Hüttmeir et al. (2009) examine the choice firms have to make between lean

practice heijunka (leanness) and just-in-sequence (responsiveness) to remain competitive.

Based on the case study of a BMW engine plant, they propose a hybrid strategy with first

using heijunka to smooth out the most extreme production values followed by JIS for the

remainder of the production.

5
According to Chavez et al. (2013), complexity of the lean practices-performance link is yet

not well understood and needs further exploration. Researchers have observed in the field that

scope and focus of lean practices rollout is determined by workers perceptions, which are

influenced by preexisting process characteristics and gender of employees (Losonci et al.,

2011). They noticed that plants with more transparent processes achieved moderate lean

transformation through work method and commitment, and plants with less transparent

processes achieved radical lean transformation through communication and belief. By

studying the impact of internal lean practices (comprised of pull-production systems, set-up

of
time reduction, just-in-time and quality management) on operational performance, Chavez et

ro
al. (2013) explain under which circumstances lean practices are more effective by considering
-p
industry clockspeed contingency. While developing a methodology to assess systemic
re
leanness of a value stream, Narayanamurthy and Gurumurthy (2016) explained that lean
lP

adoption undergoes three broad stages – lean implementation readiness, lean implementation
na

and lean implementation assessment. Value streams transition from one stage to another in

the lean adoption journey and poor maturity in the previous stage delivers inferior results in
ur

the stages to follow.

Jo

Replicating the overall success of lean practices implementation continues to remain

infeasible in practice. This is attributed to the piecemeal approach of firms as they merely

implement isolated lean practices and allow the means (the practices) to become ends in

themselves. This pattern leads to falling short of the underlying philosophy of lean - to

achieve an overall efficient and effective production system (Bortolotti et al., 2015). Driven

by this need, the objective of this research is to examine the sequence in which lean practices

should be implemented, especially in contexts where industry 4.0 technologies are

simultaneously adopted by firms to achieve superior performance. It is hoped that this

supports firms in replicating the overall success of lean practices implementation from one

6
firm to another. Table 1 summarizes the literature on the implementation sequence of LM

practices discussed in this section.

Table 1 – Literature on implementation sequence of LM practices

2.2. I4.0 adoption

The advent of I4.0 and the envisioned benefits from its adoption have motivated many

of
researchers to develop strategic guidelines and roadmaps to support companies in their digital

ro
transformation (Pessl et al., 2017). In general, those guidelines and roadmaps encompass a set
-p
of design principles and digital technologies that aim at providing advice on the proper
re
sequence for I4.0 adoption. However, some researchers (e.g. Erol et al., 2016; Ganzarain and
lP

Errasti, 2016) claim that those roadmaps should be tailored to each company’s needs, being

adapted and customized in a way that undermines a generalizable approach.

na

In this sense, a few studies have attempted to identify trends in I4.0 adoption, so that an
ur

implementation framework could be more assertively proposed. Based on a multi-method

Jo

study encompassing a literature review and case study, Mishra et al. (2018) have suggested a

conceptual I4.0 roadmap to support a sustainable growth in the industrial sector. More

recently, Frank et al. (2019) have surveyed Brazilian manufacturers in order to identify I4.0

implementation patterns and empirically validate a conceptual framework. Nevertheless,

companies’ low readiness level on I4.0 has hindered the examination of basic digital

technologies, such as big data and analytics.

Therefore, although the evidence on I4.0 adoption is prolific, most studies are either of a

conceptual/theoretical nature or applications with a very narrow perspective. Overall, the

scarcity of empirical evidence and the lack of understanding of what a full I4.0 adoption

7
actually is, may obstruct the determination of a specific pathway for I4.0. Additionally, the

required infrastructure and labor skills for I4.0 (Santos et al., 2017) do not contribute to a

faster and wider adoption across companies and socioeconomic contexts (Ghobakhloo, 2018).

This fact negatively influences the comprehension of I4.0 adoption from a system-wide

perspective, suggesting that further investigation on the topic is needed. Finally, Table 2

summarizes the literature on the implementation sequence of I4.0 technologies discussed in

this section.

of
ro
Table 2 – Literature on implementation sequence of I4.0 technologies

-p
re
2.3. Impact of LA on performance
lP

Already in the early 1990s, initial attempts for integrating automation using technology into
na

LM emerged (Kolberg et al., 2017). Robotics has been in use for at least three decades to

improve quality, performance and efficiency in manufacturing industries (Hedelind and

ur

Jackson, 2011). With the ecosystem that is currently being offered through I4.0 technologies,
Jo

lean automation is becoming more feasible and attractive for enhancing performance

(Tortorella et al., 2019a). Easy integration and relationship maintenance between the business

partners through internet and common cloud contributes to strong collaboration,

synchronization and better communication which enables effective supplier feedback

(Sanders et al., 2016). In addition, more advanced analytics and big data environments equip

machines to be self-aware and self-maintained, thereby achieving significant improvements

in their total productive and preventive maintenance (Dombrowski et al., 2017).

However, contradictory evidences on the impact of LA on performance are also found in

literature (e.g. Sanders et al., 2016; Tortorella and Fettermann, 2018; Rossini et al., 2019),

8
which call for deeper comprehension and exploration. For instance, it gets difficult to

embrace JIT production, small batch-sizes, and continuous improvement when integrating

industrial robots without a well-thought through strategy (Hedelind and Jackson, 2011),

which, in turn, can negatively impact the overall performance. Therefore, it is important to

delineate and understand the impact that the sequence in which LM and I4.0 are implemented

can have on performance. Table 3 summarizes the literature on the impact of LA on

performance discussed in this section.

of
ro
Table 3 – Literature on the impact of LA on performance

-p
re
3. Research method
lP

This study aims at identifying pathways to implement a high-performing LA. For that, we

conducted an empirical research, which is a recommended approach for exploratory studies

na

(Goodwin, 2005). Among the existing ways of data collection for empirical research
ur

purposes, the survey method is frequently adopted due to its advantages, such as high level of
Jo

representativeness, low cost, and provision of good statistical significance and standardized

stimulus to all respondents (Montgomery, 2013). The quantification of empirical evidence

gathered from respondents carefully selected is a usual approach in studies of similar nature

(e.g. Tortorella and Fettermann, 2018; Rossini et al., 2019). Therefore, we conducted a

survey-based study with practitioners so that we could answer the research question: “which

pathway of implementation of LM practices and I4.0 technologies can mature the LA

intervention to achieve superior performance improvement?”. The proposed research method

was comprised of three steps: (i) questionnaire development, (ii) data collection and sample

characterization, and (iii) data analysis. Each step is described in detail in the sections to

follow.
9
3.1. Questionnaire development

In alignment with our research question the applied questionnaire was composed by four

parts: (i) respondent information, (ii) implementation sequence of LM practices and I4.0

technologies, (iii) adoption level of those LM practices and I4.0 technologies and (iv)

perception of operational performance.

The first part collected information of respondents (e.g. roles) and their respective companies

of
(e.g. ownership, manufacturing strategy, country, size and sector). In the second part,

ro
respondents were asked to sequence the implementation order of LM practices and I4.0
-p
technologies in their companies. For that, we listed the 21 LM practices suggested by Shah
re
and Ward (2003) and the 14 I4.0 technologies listed by Tortorella and Fettermann (2018) and
lP

Rossini et al. (2019). These sets of practices and technologies were chosen since they were

consistently referred to by other studies (e.g. Dahlgaard‐Park and Pettersen, 2009; Marodin
na

et al., 2016; Pagliosa et al., 2019). Thus, to represent the LA implementation, we combined
ur

those 21 LM practices with the 14 I4.0 technologies, and randomly displayed the 35 items in
Jo

the questionnaire to avoid bias in the responses of the implementation sequence. Respondents

should then assign ‘1’ to the first practice/technology that they have implemented and then

look for another practice/technology that was subsequently adopted, assigning the equivalent

incremental number. If two or more practices/technologies were simultaneously

implemented, respondents were asked to give them all the same number. In turn,

practices/technologies that were not implemented at all should not receive any number. The

third part assessed the adoption level of those LM practices and I4.0 technologies according

to the 3-point scale proposed in Shah and Ward (2003): (1) no implementation; (2) some

implementation; (3) extensive implementation. The fourth and last part of the questionnaire

aimed at measuring respondents’ perceptions on operational performance improvement

10
during the last three years. Following Tortorella et al. (2018), we evaluated five performance

indicators (i.e. productivity, delivery service level, inventory level, workplace accidents, and

scraps and reworks). A 7-point Likert scale ranging from 1 (worsened significantly) to 7

(improved significantly) was applied in this part. We used a 7-point Likert scale since it

allows for a better reflection of a respondent’s true evaluation than a 5-point Likert scale

(Finstad, 2010). All items and measures are given in the Appendix.

of
3.2. Data collection and sample characterization

ro
To ensure the participation of appropriate respondents, we defined a few selection criteria.
-p
First, following the suggestion from Tortorella et al. (2019a), all respondents should be
re
knowledgeable about LM and I4.0 with a minimum experience of 2 years with both
lP

approaches. Second, respondents should play key roles in their companies to allow them to

conduct a wider judgement of LA implementation within their companies. In this sense, we

na

focused on either senior/middle managers, who could perceive the company as a whole, or
ur

engineers/analysts, who were directly in charge of LA implementation in their companies.

Jo

Third, because we aimed to assess LA implementation in a manufacturing environment, we

only included respondents that worked for product-oriented manufacturers. This criterion

helped to increase the likelihood of a more experienced respondents in terms of LA, as both

LM and I4.0 were initially conceived in this industrial context (Womack et al., 2007; Lasi et

al., 2014). However, no specific manufacturing sector was targeted due to the limited number

of companies in both countries (Brazil and India) that are concurrently adopting LM practices

and I4.0 technologies.

The questionnaire was first sent by e-mail in October 2019 to 255 potential respondents that

met the aforementioned criteria. We received 45 responses, from which 8 were excluded due

11
to unsatisfactory completion of the questionnaire. Then, a follow-up email was sent in

November 2019, adding a further 27 responses to our dataset from which 3 were withdrawn

due to lack of information. The final sample thus comprised 61 respondents, which results in

a response rate of 23.9%. To check for non-response bias, we analyzed differences in means

between early (n1 = 37) and late (n2 = 24) respondents through Levene's test for equality of

variances and a t-test for the equality of means (Armstrong and Overton, 1977). Results

indicated significance levels higher than 0.05, which allowed us to disregard the possibility of

differences in means and variances.

of
It is worth mentioning that the sample size of 61 respondents was below our expectations.

ro
However, the establishment of rigorous sample selection criteria, such as a minimum 2-year
-p
experience with both LM and I4.0, may have affected the number of responses in our dataset.
re
As noticed by Tortorella and Fettermann (2018), few manufacturing companies have
lP

concurrently implemented LM and I4.0 for a significant amount of time. Further, the
na

combination of both approaches become even rarer when considering the context of emerging
ur

economies, such as Brazil and India, which significantly restricts the number of respondents
Jo

that meet such criteria. Still these criteria are necessary to ensure qualified responses.

The final sample was reasonably balanced with regards to respondents’ characteristics, as

shown in Table 4. Most respondents had an Engineer or Analyst (57.4%) role within their

companies. Most companies were located in Brazil (59.0%) and had more than 500

employees (59.0%). The majority of manufacturers were national (52.5%), i.e. owned by

either Brazilian or Indian companies, and belonged to the food sector (32.8%). Regarding

their manufacturing strategies, surprisingly, most companies had either a ‘made-to-order’ or

‘engineered-to-order’ strategy, with 37.7% of respondents each.

Table 4 – Sample characteristics (n = 61)

12
To best of our knowledge, there is no measure to assess the validity of partial complete

heterogenous ranking data as collected in the second part of our questionnaire. Similar, the

third part on the adoption level does not measure any concept or construct. Only results from

the fourth part, performance improvements, were therefore checked for reliability using

Cronbach’s alpha. A value of 0.781 infers a high reliability of responses according to Meyers

et al.’s (2006) threshold of 0.6 or higher.

of
ro
3.3. Data analysis

-p
There was no significant implementation sequence for our initial analyses at the item level.
re
Hence, we looked for clusters based on the performance improvement levels, and the
lP

implementation sequence of LM practices and I4.0 technologies. Clustering tools are

designed to examine the relationships within a database to determine whether it is possible to

na

describe such data using a small number of observations of similar classes (Gordon, 1999).
ur

According to Rencher (2002), the objects within a cluster must be similar to the other inserted
Jo

into the same cluster (homogeneity), and different from other objects embedded in other

clusters (denoting heterogeneity). We performed two clustering analyses: one aiming to

identify different levels of operational performance improvement among respondents, and

another one considering the sequence of implementation of LM practices and I4.0

technologies as clustering variables.

For the first clustering analysis, we used observations related to operational performance

improvement as clustering variables. To identify the most adequate number of clusters, we

applied Ward’s hierarchical method (Rencher, 2002). Next, using k-means method, we

rearranged observations into the number of clusters previously identified (Hair et al., 2006). It

13
is worth mentioning that we performed an analysis of variance (ANOVA) as a post hoc

procedure to check for differences in means across clustering variables calculated using data

from each cluster.

We also tested for differences in frequencies of observations between clusters according to

each company characteristic; i.e. company size, manufacturing strategy and ownership. These

variables were considered as categorical since we were utilizing the dimensions obtained

from the clustering analysis for performance improvement and companies’ characteristics,

which allowed for the application of the chi-square test with contingency tables and adjusted

of
residuals. This procedure was applied to test the hypothesis that frequencies in the

ro
contingency table were independent (Tabachnick and Fidell, 2013). It allowed to verify
-p
whether clusters’ composition was associated with performance improvement or not. We
re
considered associations to be significant when the adjusted residual values were larger than
lP

|1.96| and |2.58|, which corresponds to a significance level of 0.05 and 0.01, respectively
na

(Hair et al., 2006).

ur

The second clustering analysis utilized the implementation sequence of LM practices and I4.0
Jo

technologies as clustering variables for the LA implementation sequence. This is partial

complete heterogenous ranking data. To create complete ranking data for further analysis,

whenever a practice or technology had its response empty (i.e. was not implemented), we

purposefully assigned the value of ‘35’ to its implementation sequence, since we have in total

35 practices/technologies and this would be the last possible number. There is no

recommended procedure for clustering heterogenous ranking data. Our ranking data represent

time which in turn can be represented as a geometrical distance. We therefore considered

Ward’s hierarchical method that focus on the squared Euclidean distance to be appropriate

for our data.

14
Differences in the mean implementation levels of each one of the clusters of LA

implementation were verified according to performance improvement levels (based on the

performance improvement clusters identified previously). For that, we applied One-way

ANOVAs, testing the null hypothesis that states that samples in all groups are drawn from

populations with the same mean values. The ANOVA produces an F-statistic, which is the

ratio of the variance calculated among the means to the variance within the samples. A higher

ratio therefore implies that the samples were drawn from populations with different mean

values (Howell, 2012).

of
ro
4. Results and discussion -p
re
4.1 Presentation of results – Performance Clustering
lP

Figure 1 depicts the dendogram for the clustering analysis based on the improvement level of
na

operational performance; two clusters were identified. Then, using the k-means method and

fixing k equals to two, clusters were rearranged (see Table 5), and the ANOVA results
ur

indicated that all five performance indicators presented significant differences in means (p-
Jo

values < 0.05). The 23 observations assigned to cluster 1 displayed lower mean values for all

performance indicators, suggesting that these respondents perceived a lower level of

performance improvement in their companies in the last three years. Hence, this cluster was

labeled as ‘Low Performance Improvement’ (LPI). The remaining 38 respondents grouped in

cluster 2 perceived significantly higher means, indicating that these companies had a higher

improvement level of their performance. This cluster was consequently labeled as ‘High

Performance Improvement’ (HPI).

Figure 1 – Dendrogram of operational performance improvement clusters

15
 Table 5 – ANOVA between performance improvement variables of each cluster

Table 6 shows the contingency table and chi-square results for all companies’ characteristics

(i.e. company size, manufacturing strategy and ownership) according to the perceived

performance improvement level of manufacturers. Frequencies indicated the number of

companies assigned to each cluster (LPI or HPI) that present certain characteristics; for

of
example, there are 11 companies that adopt an engineered-to-order strategy within the LPI

ro
cluster. Adjusted residual values indicated that the effect of companies’ characteristics on the

-p
perceived improvement level is less pervasive than expected. In fact, only company size

seems to be significantly associated (χ2 = 6.036; p-value < 0.05) with the improvement level
re
in operational performance. In other words, larger manufacturers (≥ 500 employees) have
lP

perceived a more prominent performance improvement in the past few years, as these
na

companies are significantly more frequent in HPI cluster than smaller ones (< 500
ur

employees). No significant association was found between the other characteristics and
Jo

performance improvement.

Table 6 - Chi-square test among contextual variables according to operational performance improvement

On the one hand, the identification of the influence of company size on operational

performance has been evidenced by many studies (e.g. Yeung, 2008; Aras et al., 2010; Hui et

al., 2013). However, indications and extension of this effect may vary (i.e. positive or

negative) depending on the performance metric that is considered. Our results suggested that

the improvement level of those five performance indicators is more likely to be higher when

16
considering large-sized companies. According to Muscalu et al. (2013) and Schreck and

Raithel (2018), although large companies usually present more complex organizational

communication channels, their operational performance control and results dissemination

among employees are generally more structured, allowing a more in-depth understanding

and, hence, accurate perception of the variation of these indicators. This might explain the

positive association between company size and the perceived performance improvement.

On the other hand, this result is somewhat convergent to the indications from Anand et al.

(2009) and Singh and Singh (2014), which have argued that the effect of contextual variables

of
on performance improvement might be less intense when companies have a structured and

ro
formal approach for continuous improvement. This suggests that the observed performance
-p
improvement is unlikely to be influenced by companies’ characteristics. Consequently,
re
performance improvements might be better explained by the practices and technologies
lP

companies have been adopting over the years. Thus, understanding the sequence and level of
na

implementation of LA may shed some light on the variation in performance among these
ur

companies. This will be discussed in Section 4.3 below.

Jo

4.2 Presentation of results – Implementation Sequence Clustering

Regarding the clustering analysis for the LM practices and I4.0 technologies, we initially

identified three clusters of LA implementation, as shown in Figure 2. As aforementioned,

whenever a practice and/or technology has not been implemented at all, respondents should

not assign any value to such item in the second part of the questionnaire. This resulted in the

implementation rate, which represented the percentage of respondents that claimed to

implement that practice/technology at a certain moment within their organizations. The

17
details of the second clustering analysis of the 35 practices and technologies are shown in

Table 7. In total three clusters were identified as follows:

• Cluster 1 (Start-up) was comprised by 10 practices and technologies that presented the

lowest mean values for the implementation sequence order and the highest mean

implementation rate (71.1%). This indicates that, in general, these practices are the first to

be adopted in a LA implementation. It is worth noticing that all 10 are LM practices from

Shah and Ward’s (2003). This suggests that these LM practices establish the fundamental

of
basis for a LA implementation; hence, this cluster was denoted as ‘Start-up’ practices.

ro
• Cluster 2 (In-transition) was composed by 5 I4.0 technologies and 6 LM practices, whose

-p
mean values for implementation sequence order varied from 15.0 to 21.4. The mean
re
implementation rate of these 11 practices and technologies was 44.6% and their mean
lP

values for implementation sequence were considered intermediate. Practices (e.g. pull

system/Kanban, self-directed teams and lot size reductions) and technologies (e.g. real-
na

time data sharing with suppliers/customers and RFID tags at products) encompassed in
ur

this cluster present a slightly higher complexity when compared to the practices in the
Jo

Start-up cluster. However, they are not cutting-edge practice/technologies in

manufacturing environments. Thus, it becomes reasonable to expect that this cluster

represents practices and technologies that support the transition of a manufacturer to a

more advanced level of LA implementation, which led us to label this cluster as ‘In-

transition’ practices/technologies.

• Cluster 3 (Advanced) encompassed the remaining 5 LM practices and 9 I4.0 technologies,

which had the highest mean values for the implementation sequence order (varying from

24.4 to 33.7). Practices and technologies from this cluster also had the lowest mean

implementation rate (31.9%), which corroborates the indication that these are typically

adopted last in a LA implementation. A possible explanation is the high-complexity and

18
 the strict requirements necessary to implement those practices/technologies. Further,

contrary to what was observed in previous clusters, this cluster is mainly comprised by

I4.0 technologies that specifically need more sophisticated infrastructure and labor skills to

work appropriately, such as cloud computing system and augmented reality. Based on

these arguments, this cluster was denoted as ‘Advanced’ practices/technologies.

Figure 2 – Dendrogram of practices/technologies based on implementation sequence order

of
ro
Table 7 – Clusters of LA practices/technologies based on implementation sequence

-p
re
Finally, Table 8 gives the results from the One-way ANOVA used to verify the differences in
lP

the mean implementation levels of each of the three clusters of LA implementation according

to performance improvement levels (LPI and HPI). Contrarily to companies’ characteristics,

na

performance improvement appears to be closely related to LA implementation level. For

ur

instance, when considering the Start-up practices/technologies, the implementation level

Jo

seems to be positively associated with performance improvement (F-value = 2.694; p-value <

0.05). In other words, companies that have been adopting these practices/technologies more

extensively are more likely to perceive larger leaps in their operational performance over the

years. As Start-up practices/technologies are usually implemented first and mainly comprised

by LM practices, this result reinforces that the establishment of a robust LA basis helps to

ensure relevant increments in performance, even when considering a medium-term

perspective (i.e. three years). Such outcome converges to indications from Kolberg et al.

(2017) and Tortorella et al. (2020b), which have emphasized that LM implementation

provides a solid process and behavioral foundation on which I4.0 technologies may build and

potentialize results.

19
A similar trend was observed with respect to In-transition practices/technologies, as its

implementation level appears to be positively related to companies’ performance

improvement level (F-value = 3.599; p-value < 0.05). Practices and technologies bundled in

this LA set may face additional challenges. According to Negrão et al. (2020), one of the

most critical moments for a lean implementation occurs after the “honeymoon” period, which

is typical of the beginner stage. After the short-term wins, companies need to perform

fundamental changes in their sociotechnical systems so that they keep evolving and

improving their processes (Narayanamurthy and Gurumurthy, 2016; Tortorella et al., 2017).

of
In other words, the successful implementation of In-transition practices/technologies usually

ro
goes beyond the technicalities, being affected by the way people behave and internalize the
-p
required sociocultural changes (Cassell et al., 2006; Bortolotti et al., 2015b). Nevertheless,
re
our results evidenced that the extensive adoption of these In-transition practices/technologies
lP

may lead to a superior performance, underpinning the assumption that they may positively
na

bridge the transition of a manufacturer from a beginner to an advanced LA transformation.

ur

In opposition, Advanced practices/technologies did not show a relationship with the same
Jo

significance level as the previous ones. No association was found between their

implementation level and the variation in operational performance.

Table 8 – One-way ANOVA results for mean implementation levels of practices/technologies according to

operational performance improvement

4.3 Discussion of results: schematic pathway development

Our results raised interesting insights that deserve further discussion. First, the fact that Start-

up and In-transition practices/technologies are indeed associated with higher performing

companies empirically confirms that LA implementation does have a positive impact on

20
operational outcomes. This finding consistently converges to Tortorella and Fetterman’s

(2018) and Rossini et al.’s (2019) works, which have verified the effect of the integration

between I4.0 and LM on companies’ performance. However, it adds to these studies as we

suggested a preferential implementation sequence (i.e. pathway) for LA, which provides

clearer guidelines for managers and academicians with respect to how to successfully adopt

these practices and technologies.

Second, it was observed that, as LA implementation advances, companies tend to move from

an exclusively LM approach to an I4.0 technologies orientation. This means that most high-

of
performing companies begin their LA implementation based on a solid understanding and

ro
adoption of LM practices. To continue progressing on their continuous improvement
-p
approach, these manufacturers start integrating I4.0 technologies of medium complexity, such
re
as ‘RFID tags at products’, ‘sensors for monitoring the production process’ and ‘machines
lP

with digital interfaces and sensors’, observing consistent and positive enhancements on their
na

performance level. Nevertheless, this integration does not exempt the need to refine LM
ur

comprehension since they keep adopting higher-complexity practices, such as ‘pull

Jo

system/kanban’ and ‘self-directed work teams’. This transient LA stage is very much aligned

with the exploitation phase suggested by Netland and Ferdows (2016), in which companies

are realizing the benefits from integrating I4.0 into LM. In-transition practices/technologies

are expected to support further developments for a full LA implementation. The final stage

for LA implementation comprehends the adoption of Advanced practices/technologies.

Contrary to Start-up, here companies may spend more efforts in adopting I4.0 technologies,

since LM practices were substantially addressed in previous stages. In this sense, Advanced

LA implementation concerns the adoption of highly complex and more infrastructure-

demanding technologies (e.g. cloud computing system, additive manufacturing, rapid

prototyping and 3D printing) that are likely to aid LM practices that emphasize the

21
improvement of flow within manufacturers (e.g. cellular manufacturing, JIT/continuous flow

production and quick changeover techniques). According to Womack and Jones (1997), an

efficient and defect-free flow of value is a key aspect of a lean system and usually the

ultimate goal of an organization, which is reasonably supported by the adoption of Advanced

practices and technologies. However, these practices and technologies did not yield the same

level of performance improvement as observed in the first two stages.

In fact, another insightful finding was the absence of a statistically significant relationship

between the adoption of Advanced practices/technologies and performance improvement. A

of
possible explanation is the law of diminishing returns, i.e. as improvement moves a

ro
manufacturing plant nearer and nearer to its operating or asset frontier more and more
-p
resources must be expended in order to achieve each additional increment of benefit
re
(Schmenner and Swink, 1998). While new technology shifts the asset frontier, there is a final
lP

frontier given by the current technological process. This highlights that it is important for
na

managers to be aware of diminishing performance returns when moving along our pathway to
ur

a high-performing LA implementation schematized in Figure 3.

Jo

Figure 3 – Schematic representation of the pathway to a high-performing LA implementation

5. Conclusions

This study aimed at investigating pathways to implement a high-performing LA. Our findings

have relevant implications for both theory and practice, which are detailed as follows.

With regards to theory, the identification of three sets of LA practices/technologies that

should be subsequently implemented suggests an interdependence and precedence

relationship between them. This time dimension is typically neglected in the literature which

22
tends to observe the implementation level of practices and technologies at one moment in

time. Such outcome indicates the existence of stages through which companies need to pass

so that a full LA implementation is achieved. Furthermore, our findings highlight that much

still needs to be unveiled, especially in terms of the impact of LA on performance

improvement as companies achieve more advanced levels of implementation.

In practical terms, most studies that investigate LA (or the integration of novel technologies

into LM) approach the topic without recommending a clear implementation sequence of

practices and technologies. Although most of these studies have suggested a positive

of
correlation between LM and I4.0 towards a successful LA implementation, to the best of our

ro
knowledge none of them have indicated such pathway. In this sense, managers and
-p
practitioners from companies undergoing a LA implementation may find here guidelines that
re
can support them to prioritize efforts and more objectively focus on the proper set of
lP

practices and technologies. This is particularly valid when companies realize their actual
na

readiness level and the next steps to continuously improve their products, processes and
ur

services are not quite evident. Meanwhile, it is important that managers are aware that there
Jo

will be diminishing performance returns as the LA transition progresses and the company

approaches the final asset frontier given by current technological possibilities.

It is worth mentioning that there are also some limitations to this study. The first comprises

our sampling process. Our indications are limited to manufacturers located in Brazil and

India. However, LA has been implemented in different industry sectors worldwide and

additional insights could probably be gained if sample size was increased and the sample

diversified. This would also allow the utilization of more sophisticated statistical techniques,

which could lead to complementary findings. Another key limitation concerns the sets of LA

practices/technologies. The actual validation of these sets still needs further empirical and

experimental analysis, which could be extensively carried out by future studies. Additionally,

23
companies might implement other practices/technologies that were not contemplated in our

study, which could potentially raise additional insights. These issues could also motivate

future studies on the topic. Finally, the transition throughout the proposed LA pathway is

subtle and not explicit between sets of practices/technologies (i.e. stages). Therefore, methods

that help companies assess their readiness level would complement our study, avoiding

jumping ahead to stages for which companies are not yet mature.

of
References

ro
Åhlström, P. (1998). Sequences in the implementation of lean production. European Management

Journal, 16(3), 327-334. -p

re
Åhlström, P., & Karlsson, C. (2000). Sequences of manufacturing improvement initiatives: the case of

delayering. International Journal of Operations & Production Management, 20(11), 1259-1277.

lP

Anand, G., Ward, P., Tatikonda, M., & Schilling, D. (2009). Dynamic capabilities through continuous
na

improvement infrastructure. Journal of Operations Management, 27(6), 444-461.

ur

Aras, G., Aybars, A., & Kutlu, O. (2010). Managing corporate performance. International Journal of

Productivity and Performance Management, 59(3), 229-254.

Jo

Armstrong, J., & Overton, S. (1977). Estimating nonresponse bias in mail surveys. Journal of Marketing

Research, 14(3), 396-402.

Bortolotti, T., Boscari, S., & Danese, P. (2015b). Successful lean implementation: Organizational culture and

soft lean practices. International Journal of Production Economics, 160, 182-201.

Bortolotti, T., Danese, P., Flynn, B., & Romano, P. (2015a). Leveraging fitness and lean bundles to build the

cumulative performance sand cone model. International Journal of Production Economics, 162, 227-241.

Browning, T., & Heath, R. (2009). Reconceptualizing the effects of lean on production costs with evidence from

the F-22 program. Journal of Operations Management, 27(1), 23-44.

Cassell, C., Worley, J., & Doolen, T. (2006). The role of communication and management support in a lean

manufacturing implementation. Management Decision, 44(2), 228-245.

24
Cezarino, L., Liboni, L., Stefanelli, N., Oliveira, B., & Stocco, L. (2019). Diving into emerging economies

bottleneck: Industry 4.0 and implications for circular economy. Management Decision, (forthcoming).

Chavez, R., Gimenez, C., Fynes, B., Wiengarten, F., & Yu, W. (2013). Internal lean practices and operational

performance. International Journal of Operations & Production Management, 33(5), 562-588.

Dahlgaard‐Park, S., & Pettersen, J. (2009). Defining lean production: some conceptual and practical issues. The

TQM Journal, 21(2), 127-142.

Dalenogare, L., Benitez, G., Ayala, N., & Frank, A. (2018). The expected contribution of Industry 4.0

technologies for industrial performance. International Journal of Production Economics, 204, 383-394.

of
Dombrowski, U., Richter, T., & Krenkel, P. (2017). Interdependencies of Industrie 4.0 & Lean production

ro
systems: A use cases analysis. Procedia Manufacturing, 11, 1061-1068.

-p
Erol, S., Schumacher, A., & Sihn, W. (2016). Strategic guidance towards Industry 4.0–a three-stage process

model. In International Conference on Competitive Manufacturing (Vol. 9, No. 1, pp. 495-501), January.
re
Finstad, K. (2010). Response interpolation and scale sensitivity: Evidence against 5-point scales. Journal of
lP

Usability Studies, 5(3), 104-110.

na

Frank, A., Dalenogare, L., & Ayala, N. (2019). Industry 4.0 technologies: Implementation patterns in

manufacturing companies. International Journal of Production Economics, 210, 15-26.

ur

Frank, A., Mendes, G., Ayala, N., & Ghezzi, A. (2019). Servitization and Industry 4.0 convergence in the digital
Jo

transformation of product firms: A business model innovation perspective. Technological Forecasting and

Social Change, 141, 341-351.

Ganzarain, J., & Errasti, N. (2016). Three stage maturity model in SME's toward industry 4.0. Journal of

Industrial Engineering and Management, 9(5), 1119-1128.

Ghobakhloo, M. (2018). The future of manufacturing industry: a strategic roadmap toward Industry 4.0. Journal

of Manufacturing Technology Management, 29(6), 910-936.

Goodwin, C. (2005). Research in Psychology: methods and design. John Wiley & Sons, Inc., New York.

Gordon, A. (1999). Classification. Chapman and Hall-CRC, London.

Hair, J., Tatham, R., Anderson, R., & Black, W. (2006). Multivariate data analysis, Pearson Prentice Hall,

Upper Saddle River, NJ.

25
Hedelind, M., & Jackson, M. (2011). How to improve the use of industrial robots in lean manufacturing

systems. Journal of Manufacturing Technology Management, 22(7), 891-905.

Holt, P. (2019). The Simplicity of Lean: Defeating Complexity, Delivering Excellence. Vakmedianet

Management B.V.

Howell, D. (2012). Statistical methods for psychology. Cengage Learning, London.

Hui, H., Wan Mohamed Radzi, C., Jenatabadi, H., Abu Kasim, F., & Radu, S. (2013). The impact of firm age

and size on the relationship among organizational innovation, learning, and performance: A moderation analysis

in Asian food manufacturing companies. Interdisciplinary Journal of Contemporary Research in Business, 5(3).

of
Hüttmeir, A., De Treville, S., Van Ackere, A., Monnier, L., & Prenninger, J. (2009). Trading off between

ro
heijunka and just-in-sequence. International Journal of Production Economics, 118(2), 501-507.

-p
Kolberg, D., Knobloch, J., & Zühlke, D. (2017). Towards a lean automation interface

workstations. International Journal of Production Research, 55(10), 2845-2856.

for
re
Lasi, H., Fettke, P., Kemper, H., Feld, T., & Hoffmann, M. (2014). Industry 4.0. Business & Information
lP

Systems Engineering, 6(4), 239-242.

na

Losonci, D., Demeter, K., & Jenei, I. (2011). Factors influencing employee perceptions in lean

transformations. International Journal of Production Economics, 131(1), 30-43.

ur

Marodin, G., Frank, A., Tortorella, G., & Saurin, T. (2016). Contextual factors and lean production
Jo

implementation in the Brazilian automotive supply chain. Supply Chain Management: An International Journal,

21(4), 417-432.

Meyers, L., Gamst, G., & Guarino, A. (2006). Applied Multivariate Research, Sage Publications, Thousand

Oaks.

Mishra, D., Roy, R., Dutta, S., Pal, S., & Chakravarty, D. (2018). A review on sensor based monitoring and

control of friction stir welding process and a roadmap to Industry 4.0. Journal of Manufacturing Processes, 36,

373-397.

Montgomery, D. (2013). Design and analysis of experiments. Wiley, New York.

Muscalu, E., Todericiu, R., & Fraticiu, L. (2013). Efficient organizational communication-A key to success.

Studies in Business and Economics, 8(2), 74-78.

26
Narayanamurthy, G., & Gurumurthy, A. (2016). Leanness assessment: a literature review. International Journal

of Operations & Production Management, 36(10), 1115-1160.

Narayanamurthy, G., & Gurumurthy, A. (2016). Systemic leanness - An index for facilitating continuous

improvement of lean implementation. Journal of Manufacturing Technology Management, 27(8), 1014-1053.

Narayanamurthy, G., Gurumurthy, A., & Moser, R. (2018). “8A” framework for value stream selection–an

empirical case study. Journal of Organizational Change Management, 31(5), 1001-1026.

Negrão, L., Lopes de Sousa Jabbour, A., Latan, H., Godinho Filho, M., Chiappetta Jabbour, C., & Ganga, G.

(2020). Lean manufacturing and business performance: testing the S-curve theory. Production Planning &

of
Control, 31(10), 771-785.

ro
Netland, T., & Ferdows, K. (2016). The S‐curve effect of lean implementation. Production and Operations

Management, 25(6), 1106-1120.

-p
Pagliosa, M., Tortorella, G., & Ferreira, J. (2019). Industry 4.0 and Lean Manufacturing: A systematic literature
re
review and future research directions. Journal of Manufacturing Technology Management (forthcoming).
lP

Pessl, E., Sorko, S., & Mayer, B. (2017). Roadmap Industry 4.0–implementation guideline for enterprises.
na

International Journal of Science, Technology and Society, 5(6), 193-202.

Rencher, A. (2002). Methods of multivariate analysis, Wiley-Interscience, New Jersey.

ur

Rossini, M., Costa, F., Tortorella, G., & Portioli-Staudacher, A. (2019). The interrelation between Industry 4.0
Jo

and lean production: an empirical study on European manufacturers. The International Journal of Advanced

Manufacturing Technology, 102(9-12), 3963-3976.

Samaranayake, P., Ramanathan, K., & Laosirihongthong, T. (2017). Implementing industry 4.0—A

technological readiness perspective. In 2017 IEEE International Conference on Industrial Engineering and

Engineering Management (IEEM) (pp. 529-533). December 14-17th.

Sanders, A., Elangeswaran, C., & Wulfsberg, J. P. (2016). Industry 4.0 implies lean manufacturing: Research

activities in industry 4.0 function as enablers for lean manufacturing. Journal of Industrial Engineering and

Management (JIEM), 9(3), 811-833.

Santos, C., Mehrsai, A., Barros, A., Araújo, M., & Ares, E. (2017). Towards Industry 4.0: an overview of

European strategic roadmaps. Procedia Manufacturing, 13, 972-979.

27
Schreck, P., & Raithel, S. (2018). Corporate social performance, firm size, and organizational visibility: Distinct

and joint effects on voluntary sustainability reporting. Business & Society, 57(4), 742-778.

Shah, R., & Ward, P. (2003). Lean manufacturing: context, practice bundles, and performance. Journal of

Operations Management, 21(2), 129-149.

Singh, J., & Singh, H. (2014). Performance enhancement of a manufacturing industry by using continuous

improvement strategies–a case study. International Journal of Productivity and Quality Management, 14(1), 36-

65.

Stock, T., Obenaus, M., Kunz, S., & Kohl, H. (2018). Industry 4.0 as enabler for a sustainable development: A

of
qualitative assessment of its ecological and social potential. Process Safety and Environmental Protection, 118,

254-267.

ro
Stone, K. (2012). Four decades of lean: a systematic literature review. International Journal of Lean Six Sigma,

3(2), 112-132.
-p
re
Tabachnick, B., & Fidell, L. (2013). Using multivariate statistics, Pearson, Upper Saddle River, NJ.
lP

Tortorella, G., & Fettermann, D. (2018). Implementation of Industry 4.0 and lean production in Brazilian
na

manufacturing companies. International Journal of Production Research, 56(8), 2975-2987.

Tortorella, G., Fettermann, D., Anzanello, M., & Sawhney, R. (2017). Lean manufacturing implementation,
ur

context and behaviors of multi-level leadership: A mixed-methods exploratory research. Journal of

Jo

Manufacturing Technology Management, 28(7), 867-891.

Tortorella, G. L., Giglio, R., & van Dun, D. (2019a). Industry 4.0 adoption as a moderator of the impact of lean

production practices on operational performance improvement. International journal of operations &

production management.

Tortorella, G., Rossini, M., Costa, F., Portioli Staudacher, A., & Sawhney, R. (2019b). A comparison on

Industry 4.0 and Lean Production between manufacturers from emerging and developed economies. Total

Quality Management & Business Excellence, 1-22.

Tortorella, G., Miorando, R., Caiado, R., Nascimento, D., & Portioli Staudacher, A. (2018). The mediating

effect of employees’ involvement on the relationship between Industry 4.0 and operational performance

improvement. Total Quality Management & Business Excellence, (forthcoming).

28
Tortorella, G., Sawhney, R., Jurburg, D., de Paula, I. C., Tlapa, D., & Thurer, M. (2020b). Towards the

proposition of a Lean Automation framework. Journal of Manufacturing Technology Management,

(forthcoming).

Tortorella, G., Vergara, A., Garza-Reyes, J., & Sawhney, R. (2020a). Organizational learning paths based upon

industry 4.0 adoption: An empirical study with Brazilian manufacturers. International Journal of Production

Economics, 219, 284-294.

Womack, J., & Jones, D. (1997). Lean thinking—banish waste and create wealth in your corporation. Journal of

the Operational Research Society, 48(11), 1148-1148.

of
Womack, J., Jones, D., & Roos, D. (2007). The machine that changed the world: The story of lean production-

Toyota's secret weapon in the global car wars that is now revolutionizing world industry. Simon and Schuster,

ro
New York.
-p
Yeung, A. (2008). Strategic supply management, quality initiatives, and organizational performance. Journal of
re
Operations Management, 26(4), 490-502.
lP

Zawadzki, P., & Żywicki, K. (2016). Smart product design and production control for effective mass

customization in the Industry 4.0 concept. Management and Production Engineering Review, 7(3), 105-112.
na
ur
Jo

29
 Appendix – Questionnaire measures
Industry 4.0 technologies (Tortorella and Fettermann, 2018; Rossini et al., 2019) Lean manufacturing practices (Shah and Ward, 2003)
Robotic stations on automated production line Bottleneck removal (production smoothing)
Highly Automated Machines Cellular manufacturing
RFID tags at products Competitive benchmarking
Sensors for monitoring the production process Continuous improvement programs
Machines with digital interfaces and sensors Cross-functional work force
Collaboration with suppliers/customers through real-time data sharing Cycle time reductions
Autonomous production processes (MES, SCADA etc.) Focused factory production
Artificial intelligence and machine learning algorithms JIT/continuous flow production
Integrated engineering systems (CAD, CAM etc.) Lot size reductions
Additive manufacturing, rapid prototyping or 3D printing Maintenance optimization
Augmented reality, 3D etc. New process equipment/technologies
Big data Planning and scheduling strategies

of
Cloud computing system Preventive maintenance
Internet of Things (IoT) Process capability measurement

ro
Operational performance improvement (Tortorella et al., 2019) Pull system/kanban
Safety (work accidents) Quality management programs
Delivery service level
Quality (scrap and rework)
-p Quick changeover techniques
Reengineered production process
re
Productivity Safety improvement programs
Inventory level Self-directed work teams
lP

Total quality management

na
ur
Jo

30
 Table 1 – Literature on implementation sequence of LM practices
Reference Research questions Methodology Findings
Zero defects and delayering are starting
principles - Requires management effort and
resources early on in the implementation.
Elimination of waste, multifunctional teams,
Whether to implement
and pull scheduling are core principles -
improvement initiatives (i.e.
Requires management effort and resources
elimination of waste, zero
Longitudinal case study of a throughout implementation.
defects, pull scheduling,
Sweden-based company by Vertical information systems and team leaders
Åhlström (1998) multifunctional teams,
spending 130 days over a period are supporting principles - Requires
delayering, team leaders,
of two and a half years. management effort and resources throughout
vertical information systems,
the whole implementation, but less than the
and continuous improvement)
core principles.
in parallel or sequentially?
Continuous improvement principle after the
base has been laid - Management devoted
effort and resources
late during implementation.
Longitudinal case study started
When during manufacturing in February 1993 and ended in
Management devoted effort and resources to
Åhlström and improvement a delayering of August 1995. Data were
delayering mostly early in the adoption

of
Karlsson (2000) the organization should take collected through three ways:
process.
place? participant observation,
interviews and documents.

ro
Develop a revised framework that
How novelty, complexity,
reconceptualizes the effect of lean on
instability, and buffering Case/field study of Lockheed
Browning and production costs and use it to develop
affect the relationship between Martin’s production system for
Heath (2009)
lean implementation and
production costs?
Is it better for a manufacturing
the F22. -p propositions about how the timing, scale, and
extent of lean implementation can regulate the
benefits of lean.
re
A hybrid approach where heijunka is used to
plant to use heijunka to Stylized simulation model with
Hüttmeir et al. smooth out the most extreme production
maximize its leanness, or to a case study of a BMW engine
(2009) values and JIS is used for the remainder of
use JIS to maximize its plant.
lP

production carried out.

responsiveness?
How intrinsic factors Intrinsic factors (commitment, belief) and
(commitment, belief) and external factors (lean work
external factors (lean work method, communication)
Losonci et al. Combination of case study and
na

method, communication) have direct impact on workers’ perceptions of

(2011) survey methodologies
affect the success of lean lean success. The effects are contingent on the
implementation from worker’s scope and focus of changes and is influenced
point of view? by process characteristics.
ur

To what extent do internal lean

practices impact on multiple
operational performance? Regression analysis on
Internal lean practices are positively related to
Jo

Chavez et al. To what extent is the empirical data gathered from

quality, delivery, flexibility and cost. Industry
(2013) relationship between internal 228 manufacturing companies
clockspeed moderates this relationship except
lean practices and multiple in the Republic of Ireland.
with cost.
operational performance
contingent upon IC?
Fitness bundles establish the foundation for
Which lean practices support Structural equation modeling on layering the development of JIT and TQM
Bortolotti et al. cumulative performance and is data gathered from 317 plants in bundles that are more specific and targeted.
(2015) there a particular sequence of three industries and ten While adapting TQM and JIT bundles to
practices that will support it? countries. firm’s own context, it has to further develop
its capabilities associated with fitness bundles.
How to conduct systemic
leanness assessment by
Narayanamurthy incorporating the interactions A scale has been developed to assist firms in
and Gurumurthy between lean elements for Graph-theoretic approach. assessing and comparing their systemic
(2016) achieving continuous leanness index.
improvement of lean
implementation?
The 8A framework is proposed
by reviewing the literature on
lean implementation case Utility of the proposed 8A framework for
studies. Single case study value stream selection was confirmed through
How to properly select the methodology has been adopted its successful application in an educational
Narayanamurthy
value stream on which LM to validate the application of 8A institute. Qualitative cross-validation and
et al. (2018)
should be implemented first? framework. A multi-criteria sensitivity analysis also confirmed the
decision-making approach has robustness of the value stream selection made
been employed for choosing the using the 8A framework.
value stream.

1
 Table 2 – Literature on implementation sequence of I4.0 technologies
Reference Research questions Methodology Findings
How to align companies’ Framework proposition based Results show a strong need for guided
Erol et al. (2016) strategy with the challenges on workshop sessions with support in developing a company-specific
imposed by I4.0? experts. Industry 4.0 vision and roadmap.
Involves industry within the The application of maturity models to the
How to address the challenges
pilot program; from the I4.0 may help organizations to integrate
regarding the concept of I4.0 and
Ganzarain and diversification and capacity this methodology into their culture. Results
the diversification methodology
Errasti (2016) assessment analysis of the show a real need for guided support in
based on the vision and strategy
company`s profile, skills and developing a company-specific I4.0 vision
of the company?
technologies that dominates. and specific project planning.
Review of some major The move towards I4.0 has presented new
How do key I4.0 technologies
European industrial and reconverted some relevant concepts;
Santos et al. (2017) and concepts have been
guidelines, roadmaps and which has partially been either substituted
addressed over time?
scientific literature. or improved by some new technologies.
Results for an Austrian company are
How does a company’s maturity A detailed theoretical and presented showing that organizational
help to identify their own targets practical perspective is given changes within this field are still a
Pessl et al. (2017)
to develop a specific I4.0 for the procedure model for bottom up driven process instead of a
implementation plan? the field of action human. management indicated holistic change
process.

of
How does I4.0 help to exchange A literature review combined
A roadmap towards achieving the goals of
Mishra et al. (2018) data efficiently for a sustainable with a case study have been
I4.0 has been proposed.
growth in the industrial sector? conducted.

ro
Systematic and content-
I4.0 is an integrative system of value
centric review of literature
What are I4.0’s key design creation that is comprised of 12 design
based on a six-stage approach
Ghobakhloo (2018) principles and technology principles and 14 technology trends. I4.0 is
trends? -p
to identify key design
principles and technology
trends of I4.0.
no longer a hype and manufacturers need
to get on board sooner rather than later.
I4.0 is related to a systemic adoption of the
re
front-end technologies, in which Smart
A survey in 92 manufacturers
What are the current I4.0 Manufacturing plays a central role.
was conducted to study the
Frank et al. (2019) technologies adoption patterns in Implementation of base technologies is
lP

implementation of these
manufacturing companies? challenging companies, since big data and
technologies.
analytics are still low implemented in the
sample studied.
na
ur
Jo

2
 Table 3 – Literature on the impact of LA on performance
Reference Research questions Methodology Findings
Case study with interviews, Differences between how Swedish and Japanese
Hedelind and observations and data companies work with industrial robotics are
How industrial robotics fits into
Jackson collection on performance highlighted. Create a guideline for how to design
LM systems?
(2011) measures and historical industrial robotic work cells that can easily be
production data. integrated into LM systems.
Bridges the gap between I4.0 and LM by
Sanders et al. How LM can be implemented
Literature review. identifying exactly which aspects of I4.0
(2016) through the technologies of I4.0?
contribute towards respective dimensions of LM.
Several I4.0 elements have been structured into
260 I4.0 use cases, technologies, systems and process related
How are I4.0 technologies and presented on the “Platform characteristics. Large interdependence between
Dombrowski
principles of LM systems I4.0” have been analyzed I4.0 technologies and lean practices were found
et al. (2017)
interdependent on each other? regarding the application of for avoidance of waste and cloud computing, zero
I4.0 elements. defect and big data, visualization and cloud
computing.
Based on the model-view-controller-pattern, an
Review of 41 methods of architecture for the cyber-physical systems to
What is the ongoing work towards
LM and a demonstration of loosely couple workstations to vendor-independent
Kolberg et al. a common, unified communication
Kanban method to evaluate third-party solutions has been introduced. This is
(2017) interface to digitize LM methods

of
the feasibility of unified expected to lower the integration efforts and
using cyber physical systems?
communication interface. thereby assist in transitioning to lean automation
solutions.

ro
Tortorella What is the relationship between
Multivariate analysis on LM practices are positively associated with I4.0
and LM practices and the
data from a survey carried technologies and their concurrent implementation
Fettermann implementation of I4.0 in Brazilian
out with 110 companies. leads to larger performance improvements.
(2018) manufacturing companies?
How does I4.0 adoption (Process-
related & product/service-related)
-p
Multivariate data analyses
Process-related technologies negatively moderate
re
moderate the relationship between including ordinary least
the effect of low setup practices on performance,
Tortorella et LM practices (pull, flow and low square hierarchical linear
whereas product/service-related technologies
al. (2019a) setup) and operational performance regression models on data
positively moderate the effect of flow practices on
improvement (safety, delivery, gathered from 147
lP

performance.
quality, productivity and inventory) manufacturing companies.
in a developing economy context?
What is the interrelation between Higher adoption levels of I4.0 may be easier to
the adoption of I4.0 technologies Multivariate analysis on achieve when lean practices are extensively
na

Rossini et al. and the implementation of lean data from a survey carried implemented in the company. When continuous
(2019) practices on the improvement level out with 108 European improvement practices are not established,
of European manufacturers’ manufacturers. companies’ readiness for adopting novel
operational performance? technologies may be lower.
ur
Jo

Table 4 – Sample characteristics (n = 61)

Company’s ownership Manufacturing sector
National 32 52.5% Food 20 32.8%
Foreigner 29 47.5% Pharmaceutical 10 16.4%
Company size Metallurgy 7 11.5%
< 500 employees 25 41.0% Equipment 6 9.8%
≥ 500 employees 36 59.0% Plastic 4 6.6%
Respondent’s role Automotive 3 4.9%
Manager or Director 8 13.1% Packaging 2 3.3%
Supervisor or Coordinator 18 29.5% Furniture 2 3.3%
Engineer or Analyst 35 57.4% Others 7 11.4%
Country Manufacturing strategy
Brazil 36 59.0% Made-to-stock 15 24.6%
India 25 41.0% Made-to-order 23 37.7%
Engineered-to-order 23 37.7%

3
 Table 5 – ANOVA between performance improvement variables of each cluster
LPI (n = 23) HPI (n = 38) ANOVA
Performance indicators
Mean Std. Dev. Mean Std. Dev. F-value
Safety (accidents) 4.70 6.55 29.67**
Delivery service level 4.39 6.29 64.40**
Quality (scrap and rework) 4.65 6.21 32.31**
Productivity 4.52 6.42 36.74**
Inventory level 4.17 5.39 6.83*
Note: * p-value < 0.05; ** p-value < 0.01.

Table 6 - Chi-square test among contextual variables according to operational performance improvement
LPI (n = 23) HPI (n = 38) Total
Pearson chi-

of
Contextual variables Adjusted Adjusted frequen
Frequency Frequency square
residual residual cy
< 500 employees 14 56.0% 2.5* 11 44.0% -2.5* 25

ro
Company size ≥ 500 employees 9 25.0% -2.5* 27 75.0% 2.5* 36 6.036**
Total frequency 23 37.7% 38 62.3% 61
Made-to-stock 4 26.7% -1.0 11 73.3% 1.0 15
Manufacturing
strategy
Made-to-order
Engineered-to-order
Total frequency
8
11
23
34.8%
47.8%
37.7%
-p
-0.4
1.3
15 65.2%
12 52.2%
38 62.3%
0.4
-1.3
23
23
61
1.865
re
National 10 31.3% -1.1 22 68.8% 1.1 32
Ownership Foreigner 13 44.8% 1.1 16 55.2% -1.1 29 1.194
Total frequency 23 37.7% 38 62.3% 61
lP

Note: * significant at 5%.

na
ur
Jo

4
 Table 7 – Clusters of LA practices/technologies based on implementation sequence
Mean Mean
Implementation
Cluster Practices/Technologies implementation implementation Denominationc
ratea
sequence order levelb
Planning and scheduling strategies 86.9% 3.0 1.60 (0.79)
Preventive maintenance 77.0% 4.6 1.68 (0.47)
Quality management programs 68.9% 6.8 1.91 (0.94)
Safety improvement programs 65.6% 8.4 1.76 (1.02)
1 Continuous improvement programs 72.1% 9.4 1.68 (0.68) Start-up
(n1 = 10) Process capability measurement 68.9% 10.6 1.56 (0.62) (71.1%)
New process equipment/technologies 65.6% 10.9 1.70 (1.09)
Cross-functional work force 59.0% 11.4 1.79 (0.98)
Cycle time reductions 70.5% 12.7 1.63 (0.64)
Bottleneck removal (production smoothing) 77.0% 13.9 1.63 (0.85)
Pull system/Kanban 37.7% 15.0 1.56 (0.56)
Focused factory production 39.3% 15.2 1.80 (0.41)
Self-directed work teams 44.3% 16.4 1.42 (0.50)
Lot size reductions 42.6% 17.2 1.41 (0.50)
Real-time data sharing with suppliers/customers 42.6% 18.2 1.35 (0.49)
2 In-transition
RFID tags at products 34.4% 19.1 1.59 (0.50)
(n2 = 11) (44.6%)
Highly Automated Machines 41.0% 19.5 1.69 (0.47)

of
Maintenance optimization 50.8% 19.7 1.41 (0.64)
Total quality management 54.1% 19.9 1.53 (0.65)
Sensors for monitoring the production process 49.2% 20.0 1.26 (0.76)

ro
Machines with digital interfaces and sensors 54.1% 21.4 1.28 (0.68)
Additive manufacturing, rapid prototyping, 3D printing 21.3% 24.4 0.96 (0.64)
Augmented reality 19.7% 25.2 0.83 (0.71)
Artificial intelligence and machine learning algorithms
Robotic stations on automated production line
Reengineered production process
-p 19.7%
23.0%
27.9%
26.0
26.9
27.7
0.70 (0.72)
0.90 (0.79)
1.14 (0.74)
re
Autonomous production processes (MES, SCADA, etc.) 34.4% 28.7 1.03 (0.86)
3 Big Data 34.4% 29.2 1.13 (0.72) Advanced
(n3 = 14) Competitive benchmarking 39.3% 29.6 1.52 (0.69) (31.9%)
lP

Internet of Things (IoT) 27.9% 30.5 1.04 (0.69)

Cellular manufacturing 26.2% 30.7 0.96 (0.76)
JIT/continuous flow production 39.3% 31.6 1.13 (0.78)
Quick changeover techniques 47.5% 32.3 1.39 (0.69)
na

Cloud computing system 45.9% 33.2 1.29 (0.77)

Integrated engineering systems (CAD, CAM, etc.) 39.3% 33.7 1.09 (0.88)
a
Notes: Rate calculated out of a sample of 61 respondents.
b
ur

Numbers within parentheses represent the standard deviation of the implementation level of each practice/technology.
c
Numbers within parentheses represent the mean implementation rate of practices and technologies of the cluster.
Jo

Table 8 – One-way ANOVA results for mean implementation levels of practices/technologies according to
operational performance improvement
LPI (n = 23) HPI (n = 38)
Practices/ Mean 95% conf. interval Mean 95% conf. interval
Std. Std. ANOVA
technologies implementation Lower Upper implementation Lower Upper
dev. dev. F-value
level bound bound level bound bound
Start-up (n = 10) 1.13 0.86 0.75 1.50 1.42 0.52 1.25 1.59 2.694*
In-transition (n = 11) 0.75 0.99 0.33 1.18 1.20 0.83 0.93 1.48 3.599*
Advanced (n = 14) 0.77 1.12 0.29 1.26 1.16 0.92 0.85 1.46 2.098
Note: * p-value < 0.05.

5
Cluster 2 Cluster 1

Figure 1 – Dendrogram of operational performance improvement clusters

of
ro
-p
re
lP
na
ur
Jo

Cluster 3 Cluster 2 Cluster 1

Figure 2 – Dendrogram of practices/technologies based on implementation sequence order

1
 of
ro
-p
re
lP

Figure 3 – Schematic representation of the pathway to a high-performing LA implementation

na
ur
Jo