Written by Henry Poskitt
In future manufacturing will require a balanced approach to machine-human interactions and detailed understanding of digitally-enhanced operator contexts. UX and Interaction Design is perfectly placed to research, test, and execute a new paradigm for work in industry.
Before we start it’s worth looking back over the history of industrial progress to better understand the upcoming transformation.
History of Industrial Revolutions
The first phase of the industrial revolution was about the replacement of craft and manual production with powered machines. This led to replacement of craftspeople with machine operators: weavers by mechanical loom operators, blacksmiths by steelworkers, and so on. The need to be close to a power source concentrated industry into factories.
The second phase was the introduction of mass production epitomised by Fordism in the early 20th century. The increase in productivity allowed for increases in wages and transformed the workforce into mass consumers. This virtuous cycle was the engine of growth in capitalist economies throughout the 20th century and undid many of the social ills of the 19th century’s industrial revolution.
The development of programmable controls for machines in the 1970s led to the third phase, one of increasing automation and the gradual reduction in repetitive work for unskilled operators.
The most recent phase has been called ‘Industry 4.0’. This is defined by the ubiquitous use of robotics, big data, and, increasingly, connected machines and logistics to create an adaptive and largely autonomous manufacturing and logistics system. The endpoint of this philosophy is the ‘Lights-Out Factory’ where manufacturing is completed without human intervention – in the dark, without the need for heating and air-conditioning. The end result is more energy efficient and has a lower carbon footprint than conventional processes.
While Industry 4.0’s promises of efficient, optimised manufacturing have been delivered – the adopters of the approach have experienced certain teething problems.
A transition from discreet machines with a human operator to connected autonomous systems of many machines adds complexity. This leads to many more points of failure, each of which must be addressed by a skilled human operator. Achieving the final stage of a completely automatic factory is often hideously complex and tends towards diminishing returns.
Even modern machines struggle to match the adaptability of a human workforce. This is less of a problem in volume manufacture, but individualised products and short runs are becoming the norm in many industries – and humans are proving hard to replace.
As the manufacturing lines have become more complex the number of specialists required to design, set-up, and maintain them has snowballed. Many of these specialists such as data analysts, network engineers, and robotics engineers are in very short supply. The unglamorous preconception of manufacturing can make these specialists hard to attract. So, while the reduction in staffing numbers has relieved one HR headache, specialisation has created a new one.
Getting the most out of new technology often requires some fundamental changes. Directly translating existing processes for new technologies often fails to take full advantage of the opportunities for broader improvements to quality, safety, productivity, or employee wellness.
Although the concept of Industry 4.0 is less than a decade old and has only been adopted in certain contexts, there is already a plan for the next industrial phase – unsurprisingly called ‘Industry 5.0’. Originating from the European Commission this seeks a more balanced approach to industrial development. Where Industry 4.0 seeks to replace humans wherever possible, in contrast Industry 5.0 envisages a collaboration between man and machine: digitally-empowered operatives and decision-supported robots.
The Role of Human-Centred Design in Industry 5.0
Human-centred design is an approach to interactive systems development that aims to make systems usable and useful by focusing on the users, their needs and requirements, and by applying human factors/ergonomics, and usability knowledge and techniques. This approach enhances effectiveness and efficiency, improves human well-being, user satisfaction, accessibility and sustainability; and counteracts possible adverse effects of use on human health, safety and performance. ISO 9241-210:2019(E)
The confluence of AI-enabled decision-making, hyper-contextual information systems, connected machines and robots with human operators is going to create a novel set of design problems. At Frontend we are already encountering the beginnings of this problem space. The solution will likely come over time using the same HCD techniques that product designers have been using for sixty years: research, resolve, test, and iterate.
Here are some insights into what we have found out so far:
There is no alternative to no-holds-barred contextual research
In manufacturing, like other complex contexts, designers need to examine the problem space in detail and from every possible aspect before starting to look at solutions. In a recent project we followed ten production lines in two countries step-by-step from goods receiving through to finished goods going out on a truck. We also looked at accounting, management information, quality control, sales and marketing, and crucially IT capability in order to gain the fullest possible understanding. We followed this with an extensive total system review of existing operator systems, parts control, and logistics. All of which was a great deal of effort, but none of it was wasted. It allowed our design team to be confident and decisive and, ultimately, to successfully make bold and impactful change.
Focus on Intent
It’s normal to be overwhelmed by the scale and complexity of manufacturing processes at first sight. Whether you are trying to understand an existing process or to create a new one it is important to not get stuck in the weeds. We have found a successful approach is to start with a clear understanding of the intent of each process or sub-process. For example: identify a faulty part and order a replacement. From this starting point our approach is to be reductive, taking away any interaction that does not serve the intent and simplifying the process. Simpler underlying processes make it easier to streamline operator interactions with efficiency and training benefits.
Robots (and other automations) are now being integrated in ever more manufacturing processes. What we have learned is that the introduction of each new and more automated component requires a reassessment of the entire process. Very often a one-for-one replacement of a manual operator with automation fails to provide the expected improvement. Some of the reasons we have seen include:
- Misunderstanding what the operator is doing in practice – in contrast to the process that was originally planned at the design stage.
- Shifting the throughput pinch-point upstream or downstream to another (often manual) process. This often caused by a sub-optimization approach where only parts of the system are considered.
- Underestimating the impact of exceptions and inconsistencies which human operators have learned to work around. It is critical that new solutions incorporate the tribal knowledge of the workforce.
- Under utilising the full range of capability provided by automated machines. They can often complete tasks simultaneously that may currently be scattered around the line.
- Very different cost-per-task compared to manual operations. For example, manual handling costs will favour more tasks at less workstations, whereas smart conveyors make moving product to a larger number of more specific stations more efficient.
The route to success is to examine the as-is situation on the ground and the capabilities of both machine and human actors and then to design a novel process that utilises the best of both.
Design for Transition
Typically clients come to our agency with a tangle of UX, software systems, and business problems; irrespective of their vertical; enterprise, healthcare, or manufacturing/logistics. Given enough time and access to the right stakeholders we can usually identify which factors are influencing end-user experience, sales, retention, costs, and so on. With more analysis it is usually possible to sketch out a future solution. Either a service, software system, or physical product that would deliver a better outcome – whatever that is.
In most cases the challenging part is figuring out a sequence of design and development steps that get from the current, less than ideal, situation to a new improved state. Big Plan Up Front all-or-nothing projects are out of favour given their tendency to collapse. So today everything has to be agile with some value delivered at each step. Which is all very well, but difficult to achieve in practice. Our approach can be described as preemptive modularity. Our goal is to deliver each module or new feature as if it is part of the future solution – with defined behaviour and interfaces to other modules. Even if it sits beside, and interacts with, legacy solutions when it is initially delivered.
Design for Continuous and Increasingly Rapid Change
If the Covid crisis has taught us anything it is that timely response to change is vital to business success. This is as true for manufacturing businesses as it is for pop-up coffee shops – just a little bit harder to achieve. For industrial solutions this means that anything which assumes a static future state will fail. Instead systems should be conceived of as dynamic frameworks/platforms that deliver transitory solutions or applications to address a current need but ready to be torn down or extended as required. Our experience is that there are a number of factors that impact the speed of response:
- Disorganised and fragmented systems are brittle. So change often results in unpredictable consequences. This makes teams resistive, cautious, and reliant on extensive testing. Each development activity must simplify and normalise the underlying process rather than adding to fragmentation and complexity.
- Scarce developer and IT resources are a pinch point in systems that require coding to effect change. The design of new additions should always include easy to use low code/no code tools for operations to maintain and reconfigure with minimal external expertise.
- Static systems tend to decay exponentially over time as bits are bolted on and patched up. New additions should be extensible by default to allow ongoing expansion without degrading performance or consistency.
The solutions are relatively obvious but must be planned from the outset as they are more or less impossible to retrofit. The modularity discussed above is critical, but this must be underpinned by careful attention to configurability and configuration tooling. Changes should be configuration-only and planned for business-level users (process engineers, shift leaders, accounts admins) rather than IT specialists. Changes should be configured and tested as close as possible to the place of work by users who understand the real-world impact.
Ubiquitous Data Collection for Continuous Improvement
Simple transactional event logs are effectively useless to gain an understanding of real world operator tasks. Because they tend to only log the eventual outcome and ignore how it was achieved. Poor data collection can sometimes also create perverse incentives and counter-productive user behaviour in order to fool the system. Our suggestion is to track user actions down to individual tasks or steps. Such logs are generally manageable and do not create a huge load on systems. Fine-grained information can be invaluable for analysis and process improvement.
What Human-Centred Design Processes Can Deliver
Total System Optimisation
- Improved human/machine interface reduces training and increases staff utilisation through greater flexibility.
- Consistent interaction and modular component-based design reduces new product introduction (NPI) cost.
- Greater productivity and optimisation as a result of detailed user task analysis.
Detailed task analysis provides a platform for transition to:
- Artificial Intelligence and Machine Vision
- Digitally enhanced operators and Augmented Reality
- Hyper contextual information systems and decision support
- Ubiquitous data collection provides visibility into manual operations.
- Simplified process design tools (IDE) and templated no-code/lo-code approach reduces the requirement for specialist skills.
As smart machines and human operators become increasingly entwined in industrial and manufacturing processes there is a need for a new pattern for human and machine interaction. Digital product designers have been defining and refining the interaction between humans and software systems for the last 30 years. In industry the next challenge for designers is to blend the flexibility of human operators with predictive analytics, decision support systems and robotics to take full advantage of the latest advances in production technologies.