Idle thoughts on modularity and loose coupling in digital media

The following is a very rough sketch followed by very rough notes on the modularity and coupling dynamics in digital products.

A map of the digital media market

Obviously, given that this map only has X and Y axes, it is literally a two-dimensional view of the digital product marketplace.


  • The bespoke versus commodity axis reflects how much scope the platform allows for the producer to create differentiated products. The producer can underuse this scope either by design, by accident, or due to a lack of capability.
  • The loose versus tight coupling axis reflects how tied the product is to the vending platform. Loose coupling means that the producer can formulate their own product strategies. Tight coupling forces them to follow the product strategies dictated by the vending platform.
  • Production costs tend to go up the less commodified a product but this generally comes hand in hand with an increased ability to make money.
  • Placement on this map does not strictly speaking dictate profitability. The producer’s overall ability to make money is more often limited by how well they can adapt to the platform they have chosen (or are forced into). Lack of profitability is caused by a product/platform mismatch. Commodity products are less profitable on a per-product basis, though, than more bespoke products, can make it up in other ways.
  • All else being equal, product businesses on the left hand side are less fragile than products on the right hand side. E.g. the reveneue stream from a subscription website is going to be less fragile than the revenue stream from a subscription app. When your product is tightly coupled to the vending platform, a simple policy change can annihilate your business.
  • DRMed media are more tightly coupled to the vending platform by definition and completely remove the producer’s ability to affect the consumer’s post-purchase behaviour.
  • ePub as specified has many more capabilities than Kindle ebooks. However, the less commodified nature of ePub over Kindle ebooks is largely theoretical because the products tend to be defined by and limited to the capabilities of the larger market (the Kindle).
  • ePub is much more tightly coupled to its platform (even when DRM-free) than websites in general because reading systems have numerous overrides, frequently modify or process the ebooks, and have undocumented or poorly documented behaviours.
  • Physical goods are beyond the remit of this map but print deserves a special mention. Much like the web stack, print is exceptionally loosely coupled with its vending platforms. Anybody can print a book. Anybody can sell a book. As a format it offers tremendous scope for design and differentiation.
  • Amazon Retail’s dominance of the print retail market is evidence that print’s capacity for differentiation is being underused. The dominant product strategy at most publishing houses favours mimicking pre-existing books over differentiation. (In fact, there is a very strong aversion to differentiation—being different—throughout the publishing industry.) There is no exclusivity, no format incompatibility, no systemic reason for the market to stay with Amazon. They are, quite simply, the simplest, quickest and easiest way to buy books that—through publisher choice—are by and large fungible. Changing that requires Amazon to suddenly become bad at selling commodity consumer goods.
  • Even when they are not cross-platform like Electron apps, desktop apps are fairly loosely coupled to their platform. Anybody can make or sell them. Installation and distribution is available to anybody that can serve up HTTP.
  • The lower half represents greater overall profits for the vending platforms as they can force the price down on commodity products and take more for themselves. In some cases, such as with Amazon Web Services, the products—open source software—are both entirely commodified and loosely coupled and so effectively free. AWS is not a vending platform for web apps or server side apps—it’s a vending platform for virtual machines and web app container engines. How the buyer uses the services it vends is not its concern, as long as it isn’t legally liable.
  • The web stack’s big advantage is that it’s both loosely coupled and offers considerable flexibility in terms of commodification and differentation. It accomodates anything from a generic WordPress blog to a large scale web service with millions of users (Facebook).
  • The inherent tension in current attempts to standardise Portable Web Publications is that some members are using it to try and move ePub over from the right half to the left while others are using it to try to expand the web stack to accomodate publishing-oriented products. These two efforts are not necessarily compatible but might well be orthogonal. That is, as long as people accept that these are different efforts with different goals requiring different pathways, they should not harm one another.
  • Ad-supported websites are completely fungible commodities, rendered undifferentiated by the standardisation of ad units and services. The modularisation of the ad market doesn’t commodify the ads—it commodifies the websites displaying the ads. This commodification is demonstrated by the fact that—in business terms—they could all be replaced with pictures of kittens without their ad networks skipping a beat. Traffic is traffic. I can’t think of any reasonable long term strategy that would make ad-supported web media a viable standalone industry. Any ad-supported web media that can be replaced with a kitten picture, will be replaced with a kitten picture.
  • All of the above applies equally to ad-supported Youtube channels. Long term, these are self-sustaining hobbies at best.
  • The difference between sponsored web media (sponsor and producer have a relationship) and ad-supported web media (advertiser and producer don’t have a relationship) is very important.
  • Similar to sponsorship is the strategy of using commodity media products to sell related paid and differentiated products. E.g. Youtube channels that build an audience and then build other products to sell to that audience. Or using a blog to promote a paid business-to-business web service.
  • Netflix (and Amazon Prime Video) specialise in delivering highly differentiated content. Due to the expense and production costs of creating these series and these movies, the producers have enormous leverage. Netflix in turn benefits from being loosely coupled itself from the platforms below it in the stack—it can always run on the web if app platforms give it trouble. Its offering is differentiated enough for platform vendors (e.g. makers of smart TVs) to make allowances for its needs. Creating its own content makes it less dependent on its providers and strengthens its bargaining position.
  • AMP is Google’s attempt to make the medium match the market. Ad-supported web media have the price-taking capability of fungible commodities (that is, they have no option but to accept the price offered) but also have the operating costs of a highly differentiated product. This is an untenable business dynamic. With AMP, Google is attempting to lower web media producer costs while at the same time increasing the performance of web media on mobile phones. It is a sensible strategy if you assume that web media companies are viable in the long term. AMP’s core problem is that it does not solve the fundamental business problem that makes these producers unviable; it just reduces the tension. The advantage of the AMP concept (using custom elements to create specialised subsets of the web stack) is that it lets web-based industries and business sectors lower their costs and increase their interoperability through ad hoc standardisation using community-managed open source code.

Moral Hazard

Vending platforms that are tightly coupled and commodify the products they vend, like Facebook, Youtube, or Amazon, have substantial built-in moral hazard. Google benefits from a certain amount of ad fraud. Amazon benefits from having a certain proportion of cheap counterfeit goods in its retail site. Facebook benefits from exaggerating video views on its platform. They get a share of the upside but are unaffected by the downsides and negative externalities. It is unrealistic to assume that amoral corporations will not take advantage of this moral hazard and shift costs and risks onto their producers instead of taking them on themselves. This can inflate profits enormously in the near term but in the long term can lead to dysfunctional markets and the platform’s destruction.


Another risk is sustainability. When the platform is built around vending low cost commodity products such as open source software or self-published ebooks, it can easily result in a market where a too small proportion of the revenue ends up in the producer’s hands. The initial burst of revenue as the market gets off the ground jumpstarts the producer economy but as the market develops and ongoing costs pile up, it becomes obvious that the revenue is nowhere near enough to sustain a majority of the producers. This can result in general market stagnation.

The only way to counter this is to pay the producers more money, either directly or through secondary markets.

Sometimes this means that product development is paid for separately. E.g. the open source software products that are the basis for AWS are often partially funded by the buyers of services from AWS separately from their purchase of the services. By separating the product business into two independent markets—development versus distribution—the industry becomes more flexible and more adaptable. The popularity of Patreon and Kickstarter suggests that this is potentially a generally viable strategy for digital media producers.