In 2018, this essay argued that digital transformation would demand new ways of measuring value. Seven years later, digital technology has reshaped industries — yet our metrics and valuation practices remain largely unchanged. This reassessment examines why that gap persists, and why it now matters more than ever.
Over the past decade, digital transformations have become ubiquitous, typically focusing first on technology adoption and second on process re-engineering. These efforts have become so widespread that, in principle, the economy as a whole should now reflect their impact. We would expect to see broad changes in how firms operate, create value, and are measured. Instead, much economic behaviour — and most of the analytical concepts used to interpret it — remains anchored in older, asset-centric logics. The absence of metrics suited to interaction-based value creation is a central reason why many firms, and the economy at large, continue to look unchanged despite unprecedented digital investment.
Looking at the emergent champions of the last decade — the platform companies, the AI-natives, and other technology-driven enterprises — it is evident that new business models and novel metrics such as participation, engagement, network size, and data value now drive the majority of aggregate stock-market valuation. Yet for firms around the median, this shift has not occurred. For many historied companies, value creation and business transformation have lagged technology adoption, and the descriptors they use to define and measure their performance remain largely unchanged.
The concept of digital transformation (DX) remains amorphic. A modernized broad definition is that DX is the process of using digital technologies to fundamentally change how an organization operates, creates value, and engages with customers, employees, and stakeholders. It goes beyond simply adopting new technology – it involves rethinking business models, processes, culture, and strategy.
Fundamentally, we argue, a core of the digital transformation is a transition from push-modal economics in which the product, production and logistics drive the business, to pull-modal economics, whereby customer engagement and demand-side activities drive critical processes and value.
At this level of abstraction these phrases risk collapsing into mere buzzwords, yet if they are taken seriously, we should now observe fundamental shifts in the premises of value creation.
From the asset driven to the algorithmic enterprise
The common understanding, which has not changed much since 2018, is that a firm is defined by its assets, and that economic value is generated from the assets in a process of production. The value, including intellectual value, is encapsulated in products, which are then transacted with customers
With digital transformation, a key change is that value creation is no longer primarily asset based. Actions based on leveraging contextual interactions within multi-party networks are far more productive than traditional asset-based transactions. Google, Airbnb, and Amazon now stand as paradigmatic examples of this logic. The focus of value creation shifts from maximizing asset use, to maximizing interactions and their contextual relevance.
The new model also fundamentally drives a change away from closed, end-to-end value chains, to instead dynamically integrate contributions from an entire network around a series of contextually co-ordinated interactions. The digital transformation is successful when it realigns the business on resolving the coordination challenges that previously would not have been manageable. For want of a better term, this becomes the algorithmic enterprise.
Network economies of scope scale geometrically with the reach of the network—commonly approximated by Metcalfe’s Law. This dynamic is far more powerful than traditional economies of scale, which grow linearly and eventually face diminishing returns. The key source of value becomes the scope (breadth, depth, etc.) of the interactions with customers, contextual understanding and learning of those interactions and the ability to match unique needs with unique solutions. This is the algorithmic business and, in a multi-sided context the algorithmic platform business.
A key consequence of a digital transformation is that it should achieve a fundamental change in the information transaction costs of the enterprise. The ability to ‘negotiate’ interactions automatically and to contextualize those interactions should improve radically; recommendation systems, dynamic pricing, node-matching, data driven modelling and multiple layers of transaction are some practical expressions of this. If the interactions are placed within a network context, as opposed to restricted to a linear chain-like structure, both the scale and scope of the interactions should expand exponentially. This is the dynamic that transforms value creation – it should be what the company is targeting, when in the DX it “rethinks business models, processes, culture, and strategy”.
The enterprise still does not operate without assets, but the assets are no longer the key driver of the value generation – it is now customer and partner reach, participation, engagement and data value. As the value creation becomes detached from the assets a corollary is that asset ownership can also be disintermediated and detached from consumption.
But, if value generation is no longer due to the assets, how should we measure it? And given the original link between Assets and Equity, is the value that stems from a transaction network even attributable to Equity?
A side note on the hyperscalers
The massive investment in data centres, GPUs, networking and energy infrastructure gives the appearance that digital value is once again rooted in physical capital. Yet these assets are not the source of value. They are the infrastructure required to scale interaction-based, algorithmic value creation. Traditional firms invest in assets to produce goods and services; hyperscalers invest in assets to enable algorithms, data flows, and learning systems.
In the business model the value arises from algorithmic matching, learning from interactions, user participation, network centrality, engagement loops and data flywheels.
Scaling these dynamics, however, requires enormous upfront capital. The capital race to build and control the infrastructure layer is not because the infrastructure itself creates value, but because it is the substrate on which learning and interaction take place.
Moreover, the economics are expected to follow network logic, not industrial logic. Value is generated through the scale, quality and learning potential of the interactions that the assets enable, and returns are compounded through non-industrial mechanisms such as network leverage, data-driven compounding and ecosystem lock-in.
The implications for the financial representation of the firm
If value is no longer generated by assets, yet equity is defined as the residual claimant on the asset base, can or should we still attribute the value of the enterprise to the equity holders?
Our understanding of the firm is obviously reflected in its financial representation. In standard discourse, the balance sheet is the core representation of the enterprise; Assets are the core of the balance sheet, the other side of the balance sheet being Equity, before any concept of debt is introduced. As Equity = Assets and Assets = Value Creation, the value is thus rightly attributed to equity, and to the equity holders.
Conversely, the objective to maximize the equity holders return (RoE) translates to an objective to maximize the return on the Assets. Most traditional financial performance indicators related to the enterprise revolve around measuring asset use. And, if measures are not directly asset driven, they are related to the cost of production (itself a function of asset use) in relation to the monetization of the product.
Furthermore, we need to note that the people who work at the company are characterized in accounting terms as a cost of operation – more precisely either a part of cost of goods sold or other operating expenses (R&D, S&M, G&A). This model of representation leads to a situation where any increase in talent is initially represented as a weakening of measured performance and there is an inherent fundamental conflict between maximizing value accrual to equity vs. value to the employees.
As the basis of the enterprise changes (the digital transformation), there needs to be a change in its financial representation. This is the need reflected in the fact that new non-asset measures account for the bulk of aggregate value generation in the stock exchange indices.
Intangible asset accounting does not resolve the underlying issues. It still struggles to conceptually distinguish spending from investment where outlays are sunk and unrecoverable; it cannot recognise latent assets such as Tesla’s accumulated driving data; and it has no real mechanism for valuing network effects such as Uber’s combined driver–customer base. Most importantly, it preserves the push-modal premise that assets sit at the core of the value-generation function. No amount of creative balance-sheet adjustment will meaningfully improve our understanding of the valuation of companies such as Meta.
The deeper challenge is that this is not merely a matter of defining new measures. Our entire accounting system rests on a financial representation that no longer fits the modern enterprise. By accounting system we mean not only the systems of indicators in use in firms, but the whole body of accounting and finance theory taught in universities, and referenced as standard by auditors, bankers and consultants.
Where do we go from here?
The algorithmic enterprise needs the correct reward functions
Just as markets are designed systems, so are enterprises; they are shaped by their objectives and these objectives are increasingly a function of metrics and algorithms. The saying “you manage what you measure” has never been more true, and while discussion may focus on the algorithms (the “managing”) it needs to focus also on the measures. Not least because companies are also social constructs and therefore those measures need to reflect a conscious choice of desirable outcomes.
Defining new metrics involves two distinct but complementary requirements.
The first set of needs are the internal metrics for the enterprise: how to drive the digital transformation, how to measure the new value creation, how to define desirable outcomes, how to set new KPIs and how to steer the new algorithms. Metrics such as engagement quality vs. quantity, learning rates, trust indicators, algorithmic fairness metrics should be obvious developments of KPIs, but it seems many companies continue to lack even the first-order level of these.
As we noted above digital transformation is about introducing automation and data driven learning to the core decision making processes, but more importantly prioritizing new processes of interactions in value networks, as opposed to optimizing or rewriting old processes of production, sourcing and logistics in value chains.
Enterprise metrics will be fundamentally correlated with the algorithmic reward functions that drive operations; get the former wrong and not only will there be a lack of guidance, but abdication of choice over desirable outcomes.
The key concept we reference here is from Tim O’Reilly, which is algorithmic failure. This refers to run-away algorithmic systems due to incorrectly coded reward functions. One example being how social media’s fake news problem is not a censorship failure, but an algorithmic failure of presenting people what they want/like to an extreme. But whereas feedback-loop misdesign becomes visible in social media, it is arguably equally prevalent across enterprise systems and governance. Algorithmic failure is not confined to social media. The adoption of AI driven decision making across enterprises will not improve the situation, and without better control over reward functions it will almost certainly make it worse.
What of the current KPI’s of the enterprise? Continuing with old control measures will continue to generate outcomes that will not be materially different than before. We contend that this is a key element of the, so called, “productivity paradox” (significant investments in technology that do not radically improve productivity). Brynjolfsson et al., as well as and Haskel & Westlake have explored this area well, but more on the macro level rather than reaching down into companies. Coyle has argued the need for metrics revision on the national accounts level, but without the connection to enterprise level finance & accounting.
Our proposition is that if you continue to manage the enterprise as before, with the same set of measures and targets as before, you will not achieve any significant change in business trajectory. To be obtuse, applying outdated metrics will result in relative or absolute capital destruction and, ultimately, business failure. To coin an aviation term this is akin to CFIT, or controlled flight into terrain.
Valuing digital transformation
The second set of metrics needed are new external measures, the tools for the market to differentiate and value enterprises. Analysts and investors that set the market valuations tend to take their cue from the companies on what to focus on, though are also limited by the disclosure of the metrics. Companies do not manage this independently, as all are bound by the disclosure and narrative of their peers.
Investors, however, need tools fit for purpose. If they cannot quantify performance, markets cannot allocate capital sensibly. The VC industry has been at the forefront of addressing this need and introducing potential new metrics. Concepts such as adoption curves, MAU/DAU, network value, engagement volumes, data-assets, and GPU gigawatts have become useful tools.
It is not always obvious, though, what the metrics actually measure, and what remains missing is the ability to structure these into a meaningful accounting framework, let alone one that would link to the finance that is still being taught in business schools. The corollary of the lack of proper financial framework is that it too easily results in failures to enforce capital discipline or accountability, particularly as concepts such as “energy and spirituality” (WeWork founder Neumann) are at times introduced into serious conversation.
As we increasingly focus on these new measures, we also see an apparent paradox that enterprise value is still attributed to the equity holders as share prices, even if the causal link to balance sheet equity is broken; the dramatic separation of the market value and book-value of equity is only one representation of this. It begs at least a latent question, that if the old mechanism and basis of attribution has broken down, is the attribution as it should be?
Public equity markets require disclosure of a significant amount of financial and operational data from listed companies. This is further expanded by the requirements for ESG data disclosure, albeit the two disclosure sets are still seldom linked. In this context, it is easy to conclude that since these disclosures are mandated, the metrics must be relevant and important. After digital transformation, or with the new models of value creation, the unfortunate fact is that much of these disclosures are meaningless, or worse. If the social function of markets is to allocate capital efficiently across the economy, should we not expect that those markets update their own functioning to reflect key changes in their listed constituents?
The high-level concern regarding the lack of appropriate metrics and financial measures, replaced by ad hoc stop-gaps, is that it raises the risk of systemic adverse consequences. In other words, market bubbles inevitably become more likely when we do not have appropriate tools to measure what lies inside them.
From network economics to new metrics for the algorithmic enterprise
A promising area of development to address some of the issues seems to lie in network economics. It is now evident that network-based systems are significantly more productive and valuable than chain-based models. Metcalfe’s Law, that the value of a network grows proportionally to the square of the number of its connected users, is a high-level understanding. In practice, this can and needs to be taken further—for instance, through metrics of network centrality, transaction velocity, or engagement richness that describe how efficiently information and value circulate within and across the platforms.
In principle, we have a theoretical ability to model, measure and give meaning to these concepts. In practice, they remain a somewhat esoteric domain far from the financial mainstream – practitioners are fluent in concepts such as operating leverage or days of inventory, but far less so in interaction density or engagement elasticity. A second order challenge is that fully understanding these dynamics will require a shift in modelling philosophy—from top-down, equilibrium-based reasoning to bottom-up, agent-based models capable of representing discontinuous and nonlinear change.
We are not currently taught to think this way. Our education remains anchored in equilibrium logic, linear causality, and asset-based representation.
Resolving this is not primarily an engineering challenge – it is a social challenge requiring evolution of an entire educational paradigm.
Conclusions for the political economy
If the incumbent enterprises that still anchor our economies struggle to execute digital transformations that meaningfully improve productivity or results, it is hardly surprising that on the aggregate level of the economy we see a commensurate lack of improvement. A lacklustre development of productivity (TFP) has been a topic of discussion particularly in economies where growth has fallen below historic trend since 2008.
Digital transformation seems to work best when starting from scratch – it produces more new winners than renews old ones. Yet the political economy struggles to cope with this. Capital is a highly mobile resource and is quick to migrate within and between economies. Labour and institutional capacity struggle to do the same. The result is a structural tension affecting social cohesion, exacerbated by what appears to many to be a disproportionate share of economic value accruing to the equity holders of a few new economic hubs
In this manner company level failures, in fundamentally changing how they operate, create value and engage, cumulate and become systemic failures. We are not suggesting that new metrics alone can resolve these structural tensions. But without a financial language capable of articulating what success looks like in the algorithmic economy, enterprises and institutions alike will remain governed by outdated measures—and therefore by outdated logics. Organisations manage what they measure; if the measures remain anchored in the past, so will the enterprises.
