What if manufacturing systems leveraged human qualities and experience instead of trying to replace them?
Bringing machines and workers closer together
As one of the initial members of the smart manufacturing venture at IXDS, I guided and conducted human-centric research in various German factories over the course of 15 months. Our contributions helped the client's IT and UX departments build symbiotic interactions between workers, software and machines.
Despite increasing automation, we could show that human qualities, experience and judgement remain key to an efficient production system and how a human-centered design process can help leverage them.
— Support the client's agile software development process with user insights to scope IT solutions that effectively distribute work between software and staff.
— Evaluate the effects of the shift in roles, tasks and processes that staff members experience when new IT solutions are rolled out.
— Help uncover hidden risks and weaknesses that might resurface once processes are automated and workers can no longer intervene.
Human input is crucial to deal with unexpected situations and monitor the quality of both goods and data. Unfortunately, as IT systems take over larger parts of decision-making, little attention is paid to the design of information that humans need to review and work with.
This can hinder the adoption of new solutions and negatively impact the production's overall performance.
I coordinated the development of the research framework we applied to analyze human-to-machine and human-to-human communication processes in multiple factories. From long-term on-site evaluations to quick prioritization workshops with stakeholders from multiple countries, I was responsible for the methodology used, analysis and synthesis of insights as well as the execution of research. Our team ranged between two to four designers and researchers.
significant productivity gains will stem from better systems not better machines
After decades of continuous improvement, significant productivity gains can no longer be easily reached by investing in better equipment or drafting better error-detection processes.
Systemic interactions within less predictable pre- and post-production processes need to designed to allow free, distributed flow of information between machines and staff along the entire supply-chain.
The misleading dichotomy of man vs. machine needs to be overcome
The influence of skilled workers is often underestimated in IT departments tasked with development of future production systems. From quality control to detection and improvised handling of discrepancies between the recorded and physical realities on the shopfloor, workers often compensate systemic irregularities.
Improving interactions and communication between workers and IT systems needs to be treated with utmost priority.
User research needs to be part of software development and deployment
People in factories are quick to abandon or work around badly designed IT systems, especially if they are not evaluated and improved iteratively. These irregularities can be hard to spot and can resurface once workers are shifted to other roles because an IT system supposedly replaces their previous tasks.
A shift in organizational mindset is needed to enable a culture where learning, testing and pivoting are part of daily business.
software can remove barriers between competing departments
Staff involved in planning and production often have different objectives in regards to the utilization of IT systems. Because planners rely on complex and precise calculations, they are quick to criticize the production workers' unwillingness to perform the steps necessary to ensure high quality of digital data (e.g. by scanning items regularly). On the shop floor staff can feel left alone with unrealistic plans that do not take day-to-day anomalies or physical reality into consideration.
Software needs to consider the tasks and contexts within both departments and support rather than replace decision-making. Routined data-collection operations should be automated.
control is less effective than organic decision-making
With automation and digitalization, many traditionally decentralized production systems are at risk of becoming more vulnerable to central planning flaws and data imprecision.
Software has to be designed with its fallibility in mind. Staff needs the ability to intervene, correct and track back decisions and operations.
Much of our project work has consisted of evaluating the the effectiveness and desirability of various IT tools and IT-mediated production processes. Let's explore how we approached these challenges that culminated in the learnings above.
Unlike research for consumer-facing services, research in factories poses unique challenges. Processes and roles are often tightly defined, making it difficult for staff members to speak about things that happen outside the scope of the rulebook. Many routines are highly habitualized, many work-arounds internalized across entire teams, departments or production sites. Inconsistent role titles, cryptic abbreviations, fragmented software tools and the 24/7 production runtime only increase the complexity.
manage complexity through a methodical approach
To spot which irregularities bear potential for improvement and which can be ignored, discipline and a methodical approach was needed. I developed a five-step research process (further below) that combined principles of Lean UX with methodology from Outcome-Driven-Innovation. This helped us to minimize ambiguity, onboard stakeholders over the course of several projects, keep learning iteratively and help us relate insights to observations in a structured manner.
Our research process followed these steps:
1. Kick-off with assumptions and hypotheses: We began each project by formulating goals, assumptions and hypotheses together with the client stakeholders.
2. Plan and develop research activities: We discussed the research set-up and identified relevant UX and technical KPIs.
3. Conduct research: We gathered input through interviews, on-site observations, diaries, insight and strategic workshops, as well as moderated usability studies and questionnaires. By cross-validating our insights through multiple methods (such as combining interviews with diaries or organizing structured feedback sessions), we were able to become more confident navigating the complex matter.
4. Synthesize insights: First we documented higher-level outcomes a staff member had during their daily work. We then analyzed and documented the process of how they reached these outcomes today, as well as their expectations along the way in a formalized manner (see images below). This helped us reveal opportunities for improvement and evaluate a newly introduced IT solution or process with less ambiguity.
5. Communicate insights well: Last but not least, we documented our recommendations in a variety of ways to make them accessible to an audience not used to perform ethnographic research.
The client's IT department has received a comprehensive list of low-hanging fruits, job stories and design principles to describe user requirements. We supported these recommendations by conceptual sketches and wireframes to make the theory more approachable and concrete.
Decision-makers have received a summary of our insights to guide optimization efforts and innovation within their respective plants.
Finally, we documented our research process on the client's intranet to show the benefits of a user-centered design process and help promote user research as an internal capability within the client's production IT.