Yeah But Like, What Actually Is The Factory Though

Yeah But Like, What Actually Is The Factory Though
Fremantle, Western Australia

Late last year I presented at a conference, and a while later I received some feedback. There were a couple of things that came up that I was sympathetic to: One was that following the (abstract) concepts was challenging, and another was that the role of the project and disciplines other than engineering didn't seem relevant to the core concern - digital factories and their delivery. The scope of that presentation was more narrow than the sum of what I've raised here to date and I hope that the big picture is clearer for those of you that attended the session. I've since moved on to discuss digital markets and what shapes them, such as societal concerns and the material world.

In my earliest days of sharing the concept of the digital factories to colleagues I learned that I should do more to differentiate the digital factory from the project as I found that people came away thinking that software developers were workers in the factory, which isn't what I was going for, although projects do have important digital infrastructure that supports digital factories, and some of this infrastructure is third-party software, and some may even be code written by developers who also contribute to digital factories, which really blurs the lines particularly for people who have no direct connection to code.

What the digital factory is in terms of primitives that are recognisable to the general public depends on the architecture for the project. Some aspects of this architecture will be more recognisable without a technical understanding than others. Let's begin with a case study from my childhood - Donkey Kong Country (DKC) for the Super Nintendo Entertainment System (SNES), released in 1994.

For the uninitiated, the SNES is a device, specifically a digital games console, and DKC is a popular game for the SNES that was available at brick-and-mortar retail. The digital factory for this experience is comprised of the device, the peripherals (e.g. the game controllers and adapters for connection to a television) and the cartridge that is the point of access for the game. Quite different from 2026 - no digital storefronts, no additional content available over the internet.

Wind the clock forward to 2010 and consider Microsoft Office, which occupies an interesting position along B2B and B2C lines. Office at this time was only offered as a standalone product and from a product perspective provided access to a portfolio of distinct units of software such as Word, Excel and Powerpoint that have more modern incarnations available today. Similar to the previous example, we have a device, a display and peripherals but even with this there are some additional complications. Unlike the SNES, personal computers are not a single product sold by a single manufacturer, they are a field of differentiated products from a range of manufacturers.

As a consumer it is also possible to mix-and-match your own compatible components, creating an enormous set of possible configurations. Nevertheless, there is a general expectation from consumers that every single one of these configurations will be capable of running Microsoft Office without defect. What makes this possible is a series of handshake agreements between an operating system (Windows in this instance), the peripherals (mouse, keyboard, and so on) or other components such as a hard disk drive, and the drivers that connect them. This is a more sophisticated point in time where the aspects of this sequence that live in digital space are all technically capable of receiving updates over the internet. It was also a transitional period for distribution more broadly: in 2010, Office was widely available at brick-and-mortar retail as well as through a digital storefront managed by Microsoft.

Let's fast forward to the present day and talk about Netflix, which was once a company that facilitated video rentals by mail in the United States and today is a subscription service for video, served over the internet. As a digital service, Netflix has achieved something a very small number of companies have been able to manage - a brand that feels like it is omnipresent in digital and physical space, from mobile phones to games consoles to desktop computers to televisions and beyond. When 'smart televisions' include one-click buttons to apps on their remotes, one of those buttons is probably Netflix.

Which makes a describing "the Netflix digital factory" a little tricky. In previous posts I've mentioned a field of truth for developers when investigating issues, and a field of realised digital products when one idealised digital product is made available on many models of device that are grouped by platform, e.g. the same app on the iPhone 14 and the iPhone 15 constitute two different realised digital products. Netflix introduces a scenario where we have one conceptually identical service that corresponds to a field of idealised digital products - Netflix on iOS, Netflix on Xbox, Netflix on a desktop web browser are all slightly different idealised digital products. It follows then that there is a field of digital factories to deliver these products, each with their own technology choices and supply chain.

Zooming out, the system architecture of Netflix contains a point of access (our apps and the devices they run on) that receives content from computer infrastructure at a large scale (e.g. the cloud). In the earliest days of the web our apps may have spoken directly to machines that Netflix owns. Over many years various layers of indirection have been introduced to address areas of interest such as cost, performance and security. One of these layers that’s particularly relevant to Netflix is the Content Distribution Network.

The idea is to hold inventory of popular content throughout the world so that a point of access can communicate with machines that are in (or around) your neighbourhood, wherever in the world you may be. For Netflix and their customers this allows video content to be served both faster and more reliably. Netflix is notable for having one of the largest CDNs in the world.

So to specifically describe the digital factory for Netflix for a particular person you would need to understand where in the world they are, the device they are using to access Netflix's content, the software they are using to access Netflix and whether the content is available on a nearby node of the CDN. If only we could say the complexity ends there. We could zoom in to particular features and consider more granular primitives, or enumerate some of the other layers. On the other side we could consider that because Netflix is so accessible across so many devices, in an entertainment room setting users of the service can be confused about which device is currently serving as the point of access, the TV itself, the Xbox, or the Apple TV unit? I can only imagine this is a regular occurrence for their support team.

Until next time.