Mar 31 2019
There is a lesson that manufacturing leaders seem determined to learn the hard way: flooding factories with new technology does not improve their performance.
Roger Smith learned it at GM in the 1980s. Elon Musk, for all his other achievements, admitted by tweet to making the same mistake at Tesla last year.
To really improve manufacturing performance, you start with, as Crispin Vincenti-Brown put it, with “what happens when the guy picks up the wrench.” You work with that person to make the work easier, faster, safer, and less prone to deviations and errors. In doing this, you apply, as needed, technology you can afford that operators can work with.
This is hard work but it pays off. It is a key lesson learned from Toyota, TPS, and many companies that implemented it under the “Lean” label. But it’s an eat-your-vegetables message. The lure of a technological shortcut is irresistible.
The latest version to hog the limelight is Industry 4.0, a German government program started in 2011. Within 8 years, Industry 4.0 has succeeded in redirecting the conversation about manufacturing from the now stale Lean into a pure technology play.
Meaning of Industry 4.0
The German word “Industrie” is not usually applied to activities other than manufacturing. In English, you talk about “the banking industry”; in German, you don’t. It’s called the “Bankbranche.” “Industrie 4.0” is more specific than “Industry 4.0.”
Beyond that, however, “4.0” is not descriptive. It’s not clear that manufacturing has neatly defined revision numbers at all, let alone the ones in the Industry 4.0 literature.
In its Spring 2019 special edition, the MIT Sloan Management Review avoids the “Industry 4.0” label, discussing instead “digital transformation.” But the digital transformation of manufacturing occurred decades ago when CNC machines replaced copy milling machines and calculators replaced slide rules.
“Digital transformation” is not descriptive of what is happening today. Fran Yáñez described the content of the Industry 4.0 “revolution” in an unstructured list of “key technologies.” We present it below with more structure, as a stack, with the addition, in blue, of items that strike us as missing from Yáñez’s list. The items in the stack need the items below and are needed by the items above.
These items are to manufacturing as instruments are to making music: possession does not equate to proficiency. A factory can have all of them and still be unable to compete.
Some of the items, in fact, require substantial knowledge and skills, so much so that it is difficult to envision how any individual could master the whole list, from advanced analytics to cybersecurity.
What makes the list less daunting is that many of the items in it are not new. MES, CMMS, and SCADA systems, for example, have existed for decades.
Advanced analytics have been used in a few high-technology industries, like semiconductors but most manufacturing managers don’t use anything beyond pie charts and stacked-bars.
We will return to Yáñez’s list in Part 2. First, we discuss a general approach to the successful deployment of this kind of technology in manufacturing.
Industry 4.0, IT, and Process Control
For lack of a better term, I like to use Information Technology (IT) as an umbrella term for all the software supporting manufacturing, except for the elements that directly drive physical changes, which is the realm of process control.
Within IT, apps interacts with human users and with other apps; process controllers, in addition, take input from sensors and issue output to actuators. The bundle marketed as Industry 4.0 spans both IT and Process Control.
Consider the following examples:
Telling an AGV to move from A to B is an IT function. Production control identifies this move as necessary and gives the order to do it. In response, the computer embedded in the AGV plots a path from A to B and drives the AGV while dodging fixed and moving obstacles. This is the job of a controller.
Failure analysis on a product unit after final test is an IT function. Making a machine perform a sequence of steps while keeping its vital signs within specified limits and taking in-situ measurements is a process control function.
The line between IT and process control is blurry but there are several reasons to draw it:
The stakes in getting the systems to work right are different. In IT, errors may lead to bad business decisions; in process control, they may destroy products and equipment, and even injure people.
The technical skills required for IT and process control are different.
IT is managed by a specialized department while control is within the purview of manufacturing or production engineering.
Dysfunctions of Manufacturing IT
The key reason given for IT to be managed centrally is that the IT apps share an infrastructure of networks, servers, workstations of different types, and a set of approved and supported software tools.
Legacies of the Mainframe Age
To a large extent, however, this organization structure is a legacy of the mainframe age. All the data processing was done in a single computer for the whole plant, interacting with dumb terminals keystroke by keystroke.
Innovation since the mainframe era has enabled the various departments to go their own way and acquire their own systems. IT departments have been resisting this evolution, leading to the paradox that employees of major corporations often have more modern tools in their private lives than at work.
Two Leaders’ Perspectives
In a 2019 interview, retired GE CEO Jeffrey Immelt said:
“Harsh though it may sound, the IT functions in manufacturing companies aren’t staffed by digital technologists. IT engineers buy hardware, outsource software development, and excel at managing projects and customizing vendor-developed software to improve operational efficiency. Reimagining products and services with proprietary software for customers requires very different capabilities.”
In a 1947 lecture (p. 12) Alan Turing had warned about the response to continuing innovation by an IT profession that didn’t exist yet and therefore could not be offended by harsher words than Immelt’s. He said:
“They may be unwilling to let their jobs be stolen from them in this way. In that case they would surround the whole of their work with mystery and make excuses, couched in well chosen gibberish, whenever any dangerous suggestions were made.”
More than 70 years later, engineers and managers in Production, Materials Management, Maintenance, or Quality may recognize in Turing’s words their experience with corporate IT.
Whenever, in operations, they try to do anything innovative with data, the IT department is in the way, blocking instead of supporting them. Second only to Accounting, corporate IT is perceived as the main obstacle to improvement.
The Perspective of Corporate IT
From the perspective of corporate IT, on the other hand, there are reasons to try and prevent a free-for-all. If every department is allowed its own information system, the plant and the company become a tower of Babel, where data generated in multiple systems cannot be cross-referenced. As of 2019, however, IT departments have been unsuccessful in this pursuit.
The Dysfunctional Top-Down Strategy
ERP systems, supposedly all-in-one solutions for manufacturing IT, have been the main applications supported by IT departments. But they have turned out to be insufficient to serve the needs of all the support functions.
The original sin of MRP/Closed-Loop MRP/MRP-II/ERP is a top-down approach that starts from planning at the top level and cascades down. The cascading down never worked because the models used at the top were simplistic and did not reflect the reality of the shop floor.
This is the opposite of the approach followed in TPS/Lean, that starts with the details of tasks at the control level where instructions directly translate into actions on machines and materials, and builds higher-level functions on this foundation.
Where ERP is used, workarounds proliferate. They are specialized systems or internally generated Excel spreadsheets that, in addition to being security hazards, have created the very towers of Babel central IT wanted to avoid.
In his famous management audit of the Tower of Babel, Frederick Brooks pointed out that the project did not fail because its goal was impossible or for lack of resources but because participants couldn’t communicate.
Focus on the Information Content
In 2019, restricting the hardware, system software and apps used in a manufacturing organization to a standard set defined and maintained by IT is neither necessary nor sufficient to ensure that the participants communicate. It is ineffective.
Don’t Standardize What Doesn’t Need To Be
It doesn’t matter whether individual computers run on Windows, Linux, OSX, iOS, or Android, anymore than it matters which brand of ballpoint pen they use. What does matter is that the devices should all be able to exchange messages, read from and write to repositories of shared data, while protecting proprietary information.
A Human Challenge
It is a human challenge, not a technical challenge. Individuals who share data — like production operators and materials handlers — must use the same formats, relationships, and versions. Departments that share data with other departments — like Production and Maintenance with equipment status — must do the same for what they share, etc.
Departments have reasons to group products in different ways. For Engineering, it may be by process and feature similarity; for Production, by volume and stability; in Sales, by market segment; In Accounting, by revenue or profit.
The different groupings, however, must be properly cross-referenced for the whole organization, so that you can retrieve the data about products made with a process P, at volume ≥V, for 16 to 24-year old females, with a value added ≥25%… The IT of most manufacturing plants today, is incapable of producing this kind of information on the fly.
Improving Existing Activities Versus Inventing New Ones
Every new generation of information technology has created many more opportunities than the manufacturing industry has been able to use, lagging behind, in particular, financial services. It should be no surprise to see it happen with the latest.
According to Immelt, “Many CEOs miss the fact that a digital transformation isn’t the same as the digitalization of an existing business.”
Referring to the IoT or machine learning as “digital transformation” is like calling a car a “horseless carriage.” While using quaint, archaic language, however, Immelt is warning against using the new technology just to improve the efficiency of current practices and advocates taking it as an opportunity to change strategy. This what Marshall McLuhan called the horseless carriage syndrome.
Immelt points out that the “digital transformation” can “alter what a manufacturing company sells” but this is true of any innovation. One year into an automotive client’s Lean implementation, working with the management of the plant to take stock of what had been accomplished, I was surprised to find that a Sales manager was the most enthusiastic.
In this business, you start by providing sample quantities of about 50 units of a product a customer wants. Happy customers follow up with orders in the hundreds of thousands per year for the next several years.
“Before Lean,” the manager said, “production was always too busy to bother with samples. Now, when I ask for a sample, they are somehow able to work it into the schedule. Orders for my line of products have doubled in the past 12 months.”
They had changed line designs and production control methods but the changes had not involved any IT or automatic control system.
A Bottom-Up Approach
If the top-down, ERP-based approach does not work, what is the bottom-up alternative?
Don’t Wait For The New System
A common, mistaken belief is that there is no point in trying to improve the performance of legacy systems. A new all-in-one that will be implemented two years from now will solve all the problems, and the supplier’s consultants will work with you to reengineer your business model and fit the system’s capabilities… This is the next iteration of the top-down strategy and not likely to deliver on its promises.
Apply Continuous Improvement to IT and Process Control
Instead, the whole range of IT and process control should be the object of continuous improvement. Many of the “Industry 4.0 enablers” can be retrofitted to legacy systems, yielding benefits in months rather than years, at a much lower cost than replacing the legacy systems.
In fact, some of the enablers, like Data Warehouses, are specifically intended to wrangle value out of the data in legacy systems.
As in production, this continuous improvement effort develops skills that make the organization a savvier buyer when actually replacing the legacy systems or designing systems for new plants/lines.
Leverage Existing Talent
As for production lines, part of the talent needed to pursue continuous improvement in IT and process control is already present in most manufacturing organizations, in employees who like to tinker with software. Left on their own, they are the ones producing the problematic Excel spreadsheets discussed above.
Additional talent is needed to provide the tinkerers with better tools, train them, and organize their efforts so that their contributions coalesce into a coherent system. Recruiting this talent may be the most difficult challenge. As Immelt points out, manufacturing companies “aren’t considered employers of choice by software engineers.”
The Industry 4.0 literature is long on technology but short on management enablers.