Q: How to combine organic Lean with top-down digitalization initiatives?
This question comes up very often in discussions in my Leading the Factory of the Future Masterclass considering the link between deeply advanced Lean or Six Sigma, ingrained within working routines and operations and digital. In my experience and what I saw they are highly compatible. They are similar because if you think of the improvement routines: define, measure, improve and analyse that control cycles of Lean Six Sigma, they can be perfectly executed using data. The routine is the same, black belts have statistical training, therefore, can at least interact with a more professional and sophisticated statistician. They are used to experimenting and stable implementation within an organization.
What are the differences? If you stick to the mantra of Lean Six Sigma, improvement initiatives start from ideas that are generated by workers. And in this case workers may have problems but the intuitional solution comes through that analysis, especially when you use machine learning or neural networks. They can come up with things that people didn't think of. For example in the chemical industry, there are cases where there is abnormal productivity of the reactor. They run neural network data on the process parameter data set and they find new ways of understanding the chemistry of the reaction they didn't have before.
Lean was usually very bottom-up and it was difficult to prove a global impact of Lean across the organization because the scalability was something that was not totally clear. Digital companies are also creating infrastructure in their low-level operations in the plants, in the offices and the factories, but they also have a central digital unit and the importance of this central digital unit is to govern this change, ensuring greater scalability. In this sense, because we are talking about data and algorithms with digital, we have greater oversight over what happens across the company and therefore a greater potential for making a sizable impact on business KPIs.
Q: How do we distinguish between a digital roadmap versus a technology roadmap? How do we define digital versus technology implementation?
Sometimes we focus too much on technology and of course when you think about the technology, we are thinking about defining data structures across different plants or different machines and you either want to host it on the cloud or locally, you consider different vendors. However, if think about organizing the transition we have to go back to the basics. What does it mean? Digital transformation is a strategic change in an organization. And changing an organization when you change its strategy means changing four things: structure, process and systems, incentive and rewards and people and culture. When we talk about technology we are just talking about systems. When you change the system, it's time to change the process. But then when you think about governing the transition, what about structure? The structure is where you allocate responsibilities and one of the fundamental decisions here is how do you divide responsibility for the digital transformation? You may have a unit in the headquarter or some hub of competence and then you need to decide what you leave to the plants.
The message of Lean remains valid. You have to think through all the stages of the transformation, namely groundbreaking, experimenting, piloting and scaling and divide roles and responsibilities: what does the headquarter do, what is the plant responsible for? Then think about the process, how do they interact with each other? You have vertical interaction as well as interaction across plants.
Deciding on the process also depends on the problem that you have. Imagine that I have a company with many packaging units. I develop centrally artificial intelligence technology for quality control on the packaging. Carton boxes are similar across many packaging lines. So in this case you want to centralize governance. What about the situation where plants are more idiosyncratic, where the mother plants are different from each other? Then you have to think about scalability differently. It's not really about copying the algorithm for plant A plant and adopting it in plant B, it’s more about formalizing how plant A went about a problem so that plan B can go about the problem themselves.
Then you have incentives. If you want to have scalable solutions, that means if, for example, plant A develops a solution that is also usable by plant B, you know that the more usable the solution the more difficult it is to develop it. What is the incentive for the plant manager in plant A to develop something that other plants can use? When you roadmap technology, you have to consider all these aspects: structure, processes and incentives.
Q: How do you manage the daily workload with reinventing the business at the same time?
That is an evergreen of improvement. Normally companies, after the crisis downsize the staff, reducing head count to the bare minimum in many areas and people are already running at more than 100 per cent capacity. The question is how do you free resources? Companies are putting real money behind creating digital direction, creating a budget for digital innovation but how do you make a case for the resources?
When you think about doing digital initiatives as pilots, it is very demanding because you are changing something, let's say in a production line or the maintenance department, real operations. You're making the change to see what happens and this takes significant time and is not always easy to do especially when you don't know whether this is going to be beneficial.
With experiments, on the other hand, the idea is that you engage in small efforts that last maximum few weeks or very few months with a very small amount of people. Typically, you can leverage external resources to see whether something is possible.
You can also put forward a proposal for a practical recommendation partner, such as an engineering school or something related to empirical research in production and use these external resources to get a first sense of the possible benefit. Then use these to make a case to ask for more resources if this is the case or to grant access to available resources if the problem is getting an approval.
Q: How do we measure the success of a digital transformation journey? How do we ensure it is sustainable, not a one-time exercise?
When you talk about measuring the success of a digital transformation initiative, if you measure it only at the end of the journey, you will never get there. You have to slice Key performance indicators across the four stages identified earlier: groundbreaking, experimenting, piloting and scaling. During the first stage of groundbreaking, the possible performance indicators are degree of speed over-accumulation of the data that you care about, data quality, data Integrity, safety - typical data engineering metrics, but with an eye on how you're going to use them later. In phase two, pilots, the benefit of KPIs is where you have a potential for improvement and how much. Maybe you thought that you could predict the failure of equipment but after doing a few pilots you say that this is much more complicated than you expected. At this point, maybe it's bad news, but what you learn, is that you’re better off focusing on something else. When you find something that works, you know that on paper, what comes out of simulation with the database could increase machine availability by 12 per cent. In the second stage, the KPIs are really what can work and what can't and for things that can work, how much and what the potential can be. In phase three, pilots, it is all about quality problems in the details, understanding where are the organizational tweaks that may become critical to ensure the success in the implementation of an algorithm. In stage four, we measure two things: overall KPIs for the company, but also adoption. The message, therefore, is that when you stage the change, for each stage you need these different KPIs.
Q: What are the platforms that you would recommend to support your digital transformation and what is the recommended data strategy?
The fundamental thing is, if you want to have a platform that allows, for example, to share data across plants, you have to understand what is the level of homogeneity between your plants. If you have plants that have the same type of equipment, of the same generation from the same vendors or equivalent vendors, then they would naturally create similar data. Then you need to decide whether you want to save all data or only a subset. It’s important to conduct experiments to understand which data is needed and with which frequency and then based on that, you define the overall data lake requirement.
Another approach is to have a minimum subset of common data so that plants can experiment with this data trying to tweak it to fix their specific activity. In this case, the question is, is there a common subset of data that can be shared? If the plants are heterogeneous with each other, then you need to know how much broadband you need for this platform? Maybe it’s just a tool for central control and therefore you probably have some set up there. But when you get into the details, it becomes a real engineering and data security problem.
Q: Could you share any failures you've seen in implementing a digital strategy and the lessons learned?
Typical failure that I see frequently is calling the vendors too early because you’re calling the vendors or large consulting corporations before defining what value you think those digital solutions will add to your processes. The risk is that they will just sell you what they have and since you didn't do the homework before, you may fail by committing to buy resources and technology that you do not necessarily need that much. To avoid this failure, an experimenting phase is important as it allows you to paint a bigger picture of what you want the technology to do for you and that picture must go into specifications of the contract with the vendors.
Another mistake I see often is staffing the analytics team before doing the screening internally for people with a passion and interest in analytics. Your people are more likely to be thinking about their next career move if you choose to ignore them, and you failed to capitalise on experience with digitalisation. When you create these teams, it’s important to make sure that you have true believers there that are eager to do that. The second important thing is to think about a career path for people that are on the analytics team. If they stay there, are they doomed to stay in an office in a peripheric plant or is there a career progression that is clearly outlined?
Finally, many companies create their digital roadmaps without benchmarking against competitors and industry. Something we do a lot at Leading the Factory of the Future Masterclass is seeing what other colleagues in the industry are doing, getting inspired by specific decisions, organisational setups and strategies.