More...More...More...More...More...More...Scalability

The ultimate goal of the business strategy is to ensure that the business can be scaled up to the level which matches the business objectives. This is usually a step-by-step process, involving exploration, acquiring early adopters, and then continuously growing the business. Which of the methods that have worked for successfully scaling purely digital businesses can be adopted by AIoT-enabled businesses? What are the pitfalls of scaling up a digital / physical business?

Clearly define your focus areas

Clearly define focus areas

Ensure AIoT product / market fit

A key prerequisite for successfully scaling a high-tech business is to constantly focus on the product - market fit. This means that the product -- including the user experience (UX), the feature set and the value proposition -- must meet the undeserved needs of the target customers. As we have just discussed, the target customers will also change over time, so the organization must be able to react to the changing needs.

From an AIoT point of view, we actually have to differentiate between the product and solution perspective -- smart, connected products will utilize the full breath of UX-related technologies -- including web technologies and mobile devices, but potentially also HMI embedded on the physical product. For smart, connected solutions developed with a focus on improving the operations of existing assets, the focus will often be more on utilizing basic web technologies for intranet-type applications. UX plays a role here as well, but will often not be as important compared to the product side. Take, for example, the escalator monitoring example from earlier. If the railway company is only making this available to a small set of technical operators, a simple UX will be sufficient. This would obviously be different for any operations support apps that are made available to a wider internal audience, such as train conductors. In this case an investment in a better UX (e.g., using a smartphone app) will be justified.

The feature set for the smart, connected product will include both physical and digital features, the latter enabled by IoT-connectivity, software and AI. The Digital Equipment Operator, on the other hand, will usually not be able to change the features of the existing physical assets, so the focus here is on digital features.

In "How Smart, Connected Products Are Transforming Competition"[1], Michael Porter and Jim Heppelmann describe four key capabilities of smart, connected products: monitoring, control, optimization, and autonomy. In the context of our discussion, the new products and services provided by the Digital OEM will probably most benefit from control and autonomy, while the Digital Equipment Operator will utilize the monitoring and optimization capabilities for his solution.

In terms of the target customers, the Digital OEM will usually address either a B2C or a B2B market, while the solutions for the Digital Equipment Operator will more often address the internal business units responsible for the physical assets.

Digital OEMs operating in a B2C market will usually address needs such as convenience and offering cool, new features. For those in B2B markets, customers are more likely looking for efficiency improvements and cost reductions. The Digital Equipment Operator, on the other hand, will focus on operations effectiveness (OEE).

AIoT Product/Solution - Market Fit

Ensure efficient exploration

Efficient AIoT exploration

Know how to cross the AIoT chasm

The business book classic "Crossing the Chasm" by Geoffrey Moore[2] describes the challenges of marketing high tech products, especially focusing on the chasm between the early adopters of a product and the mainstream early majority. This concept is especially important from the point of view of the Digital OEM.

Throughout the life-cycle of the product, he will face a number of different challenges, some of which will be very specific to AIoT. For example, in the early stage when addressing innovators, a challenge is to actually create a small series of physical products that appeal to innovators in combination with the digital features. Especially if AI is heavily used, this can be challenging in the early phase of the product life-cycle, because in this phase few reference data will be available for training the AI models. For asset intelligence enabled by AI, this will probably mean that simulation and other techniques will have to be applied. For any swarm intelligence required by the product, this will be even more challenging because the "swarm" of products in the field actually creating data will be relatively small.

When moving on to the next phase, serving the early adopters, an AIoT-enabled product will have to make difficult decisions about the MVP or baseline of the physical side of the product because this will be very difficult to change after the Start of Production. Another key point will be a strong UX to appeal to early adopters: something that start-ups tend to be better up than incumbents.

Finally, when crossing the chasm to the early majority and realizing significant growth, it will be vital to establish cost-efficient, high-quality product manufacturing. Scaling the physical side of the product at this point will most likely be more challenging than scaling the digital side of it. In addition, this market will not be a pull-market, so it will require excellence in sales and marketing.

Finally, manufacturing-centric companies often struggle with the fact that the product will have to be continuously improved to stay attractive to the users. This means not only the software side of things but also the continuous retraining of the AI models used.

Crossing the AIoT Chasm

Gabriel Wetzel, CEO of Robert Bosch Smart Home: "A key challenge are the often very high expectations, which don't anticipate the ‘trough of disillusionment’ which you usually have to cross before you will see new business at scale. You have to make sure to get through this, and not lose management support on the way."

Continuously improve commercialization

AIoT Commercialization & Continuous Improvement

Understand implications of AIoT Short Tail vs. Long Tail

A good way of looking at the scalability of the opportunities presented by AIoT is by categorizing them in the short tail vs. the long tail of AIoT: the AIoT short tail includes a relatively small number of opportunities with a high impact and thus high potential for scaling them. This usually means a high level of productization and a strong Go-to-Market focus, which requires a Digital OEM organization. The AIoT long tail, on the other hand, represents a large number of opportunities where each individual opportunity is relatively small. However, together these long tailed opportunities also represent a very significant business opportunity, provided an organization is able to harvest these smaller opportunities in an efficient way. This usually requires a "harvesting" type of organization (for internal opportunities), or a platform approach, as described earlier.

A good example for an organization that is focusing on harvesting a large number of small opportunities presented by AIoT is described in the "AIoT in High-Volume Manufacturing Network" case study in Part IV of the AIoT Playbook. This case study describes how Bosch Chassis Control Systems have built up a platform and global AIoT Center of Excellence to work closely with a global network of over twenty high-volume factories. This group is managing a portfolio of AIoT-enabled production optimization projects in different areas, but usually with a strong focus on OEE improvements for the factories. This is a great example of a "harvesting" type of organization that is required to realize the opportunities presented by the AIoT long tail in such an environment. Executing this at scale for over thirty factories requires a careful balancing between a centralized expert team and working with experts in the field who understand the individual opportunities.

When looking at scalability, it is important to understand which end -- the short tail or the long tail -- of AIoT one is working on, and what type of organization is required to be successful here.

The Long Tail of AIoT

Gabriel Wetzel, CEO of Robert Bosch Smart Home: "The short-tail opportunities will often be addressed by other market players as well. This means that investment size and time-to-market are absolutely critical. The long tail requires many custom solutions. You should not underestimate the required resources, the domain-specific skills and the market access. Not all of these can be easily scaled-up. Of course this can be addressed by a top-in-class partner management: but don't forget to budget for it!"

Ensure organizational scalability

Ensuring organizational scalability is another key success factor for smart, connected products. How can an organization successfully grow and evolve alongside the product as it matures from idea to large-scale business? DevOps mandates that an IT organization combine development and operations from the beginning, iterating together continuously through the build and improvement cycles. However, in an organization that must combine IT development and operations capabilities with physical product engineering and manufacturing capabilities, this will not be as straightforward.

Dattatri Salagame is the CEO of Bosch Engineering and Business Solutions. In the following, he will discuss the issues related to scaling up and evolving an organization for smart, connected products.

Dirk Slama: What is your take on the organization we need to build and sell smart, connected products?

Dattatri Salagame: We believe that for creating a digital product, you need somebody who has an idea, who owns the market, who understands the nuances of scaling in the market -- combined with the technology of product engineering. This handshake has to be very firm, and it has to be nurtured during the course of the product evolution. So that's the reason we believe that somebody who owns the market, somebody who has a product idea, and somebody who wants to change their revenue pattern has to be the owners of the product. In addition, we need a holistic look at how to engineer the product and how to manage the product lifecycle. The lifecycle of the product undergoes a huge transition in the course of its evolution. This transition has to be managed while the business owner is able to do his experimentation and validation in the market for scalability.

Dirk: And what are the required organizational capabilities during these different phases?

Dattatri: In the beginning you need a gang of hackers. I call them a gang of hackers because in the MVP (Minimal Viable Product) stage, you hack the solution; you are not really worried about the reliability of the product. You are worried about the feasibility of the product. So you need the gang of hackers who can quickly hack a solution in a high-fidelity experimental mode. Once you have confirmed the feasibility and you received positive customer feedback, then you need to transition into a much more rigorous, disciplined product engineering process. And this is where a multidisciplinary team comes in. You need to have the competency of classical product engineering. You need to have mechanical engineering, electrical, electronics, power management, and communication. Also, you need people who understand data, mathematical modeling of the data to mimic the physics and chemistry of the product and to create AI models, and to be able to mature the product in layers. The product matures in layers, which is important. The electronics layer, the communication layer, the network service provider, then the data, then the AI and then the reliability. So these layers mature at a different velocity, at different points of time, so you need a team which is much more disciplined in this phase. Finally, when we release the product, and we have crossed the initial phases of scaling, then you need an organization to support the product introduction. This is a game of having good suppliers to be able to manage the scale and to provide high-speed DevOps as the backbone of the digital services. So you need these three flavors of the team during the course of the project.

Dirk: Any recommendation on how to organize this?

Dattatri: Everybody is talking about digital transformation. And this is it: we need to transform existing, traditional, very engineering and manufacturing-oriented organizations so that they play together with the more agile AI and software organizations to support smart, connected products. Organizations must recognize the fact that connected product engineering is multi-velocity, multilayered, multithreaded product engineering, and multidisciplinary product engineering; hence, we need many more fluid transitions in different phases into different formats. Organizations need to be aware of this transition, both in terms of speed and content and the setup of the organization. This is extremely important. Otherwise, they risk falling through between the transitions. And this is what we call the “valleys of death”. Because if you look at connected products, seven or eight out of ten products do not actually pass in flying colors. Therefore organizational agility and the ability to transform the organization along the way is an important part of the ability to "cross the chasm", as you have introduced it earlier.

AIoT Organizational Evolution

Deal with repeatability, capacity and marginal costs

Digital businesses are seen as potentially highly scalable because their digital offerings are highly standardized and easily repeatable at very low extra cost. Physical products, on the other hand, can be much harder to scale, because scale effects in manufacturing often only apply when talking about extremely high production numbers. Even in this case, the marginal costs will not be reduced to a level as we are seeing this in the case of digital businesses.

For the Digital OEM, this means that their focus usually needs to be on creating highly standardized physical products, because any increase in variants and added complexity can potentially have a negative impact on scalability. Ideally, differentiation through product variations should be mainly focused on the software/AI side. An interesting example in this context is the Seat Heating-on-demand case, which is introduced in the product operations section: instead of having cars manufactured with individual seat heating configurations, all cars come with the same physical equipment and the configuration is done later on-demand by the customer. Of course this type of business case requires careful calculation of the marginal production costs vs. the downstream revenue opportunities over the life-cycle of the car.

For the Digital Equipment Operator, the topic of repeatability and capacity is also important. This links closely back to the long-tail discussion from earlier on: if the benefits of the individual AIoT-enabled solutions are only relatively small in comparison, then ensuring repeatability on some level is key. In the "Predictive maintenance for hydraulic systems" case study, Bosch Rexroth used AIoT to enable predictive maintenance for hydraulic components. However, since each customer installation uses the hydraulics components in a different way, AI algorithms have to be adapted individually for the customer. Bosch Rexroth has established a service offering that maximizes repeatability by standardizing the sensor packs, and establishing a standardized process for the customization of AI for individual customers. In this way, the predictive maintenance service offering is competitive, despite of its positioning on the AIoT long tail.

References

  1. How Smart, Connected Products Are Transforming Competition, Michael Porter and Jim Heppelmann, 2014, Harvard Business Review
  2. Crossing the Chasm, Geoffrey Moore, 1991, HarperBusiness