Transforming Products and Services

Three primary models characterize the digital transformation of products and services. Digitally enhanced products augment physical artifacts with digital capabilities — automobiles with telematics and satellite navigation transform a mechanical mode of transport into a connected, data-generating node. Digitised products convert the core product entirely into a digital format, such as replacing physical print media with online publishing, drastically reducing marginal distribution costs and enabling global scalability. New business models leverage digital infrastructure to fundamentally alter the transactional nature of the market, shifting from outright asset ownership to subscription-based access — the servitisation of software and mobility.

Adapting to Digital Services, Customisation, and Personalisation

Informationalisation is the process by which physical products are embedded with sensors, connectivity, and software that allow them to generate, process, and act upon data. The perceived utility becomes $U = V_0 + \beta I$, where $V_0$ is base utility, $I$ is the stream of actionable insights, and $\beta > 0$ reflects the user's valuation of information. Products that were once inert — from toothbrushes to running shoes — now generate usage data, blurring the line between product and service.

Mass customisation allows users to select from predefined menus of options. Deep personalisation uses algorithmic inference and continuous data harvesting to dynamically tailor the product experience without requiring explicit, manual configuration. The optimal level of personalisation solves:

Where $\theta \in [0,1]$ is the degree of personalisation, $\delta_i$ is the consumer's idiosyncratic preference parameter, and $s_i$ is the observed signal. The optimal flat degree of personalisation is $\bar{\theta} = \frac{\lambda^2}{k} \sigma_{\delta}^2$, where $\lambda$ is the signal-to-noise ratio. The extent of personalisation rises with data signal quality and consumer preference heterogeneity.

The firm's optimization balances the marginal gross benefit of personalisation against the marginal cost of computation plus the marginal privacy cost incurred by the consumer:

If consumers become highly privacy-sensitive, the economically optimal level of deep personalisation decreases, forcing firms to innovate in privacy-preserving technologies like federated learning to maintain the value of personalisation without increasing data collection.

Providing Information, Informationalisation, and Being Digital

When the cost of sensors and data transmission plummets, the optimal level of data collection rises, making informationalisation an increasingly pervasive strategy. The transition from physical goods to digitally enhanced services shifts the primary locus of value from the physical hardware to the continuous stream of insights and services it enables.

Sticky Ecosystems

Sticky ecosystems are interconnected networks of hardware, software, and services designed to maximize user retention through high switching costs. The switching cost $S$ is decomposed into:

Where $s_i$ is device replacement friction, $L(n)$ is the loss of interoperability (the "walled garden" penalty), and $\Delta P$ is the loss of personalized algorithmic efficiency built up over time. Because $L(n)$ and $\Delta P$ increase non-linearly with the depth of the user's integration into the ecosystem, the incumbent firm possesses significant pricing power.

The consumer stays in ecosystem $A$ if $U_A \ge U_B - S$, or equivalently $V_A - V_B + \alpha(n_A - n_B) \ge -S$. Because $S$ is typically large and positive, the inequality holds even when $V_B$ significantly exceeds $V_A$. The ecosystem need not be the absolute best in isolated feature-by-feature comparisons; its stickiness arises from the switching cost gap.

Governance challenges include data portability rights that directly attack the switching cost $S$ by enabling consumers to transfer personal data to competitors, and anti-competitive regulatory interventions such as the EU's Digital Markets Act (DMA) that mandate interoperability.

From Atoms to Pixels and Reusing Infrastructure

Digital goods are characterized by non-rivalry, near-zero marginal cost of reproduction, and high fixed-to-variable cost ratios: $C(q) \approx F + \varepsilon \cdot q$ where $\varepsilon \to 0$. This asymptotic collapse of marginal cost fundamentally destabilizes classical pricing equilibria and compels firms to redesign value capture mechanisms around strategic reuse of infrastructure and data.

Digital platforms accumulate infrastructural assets — cloud compute, identity graphs, recommendation engines, logistics APIs — that can be recombinantly deployed across product lines. Digital infrastructures exhibit strong subadditivity: $C(X, Y) < C(X) + C(Y)$, because shared data pipelines, authentication layers, and analytics engines serve multiple value propositions simultaneously.

Reusing Data and Extrapolating Data

Data is a non-depleting input: its use in one context does not preclude use in another, and its marginal value often increases with scale due to learning effects. If a model's predictive accuracy is $A(n) = A_{\max}(1 - e^{-\lambda n})$ where $n$ is the volume of training observations, then every additional user interaction raises the value of the infrastructure for all subsequent users — a dynamic underpinning the business models of digital platforms.