Why Scalable Software Still Diffuses Unevenly
The counterfactual assumes that firms can recognize a valuable software opportunity, pay to adopt it, and reuse it across their establishments. The OECD's study of UK SMEs shows why those assumptions are demanding in practice. The subject here is the gap between technological scalability and realized adoption: a software capability may be reusable in principle while remaining too costly, complex, uncertain, or poorly supported for many firms. Understanding that gap helps identify which firms can reach the scaling mechanism and which remain outside it.
Potential scaling is not realized adoption
The UK evidence begins with a maturity gradient. Businesses adopt mature digital technologies such as cloud computing and data analytics more readily than advanced technologies such as artificial intelligence, robotics, and automation. SMEs have largely narrowed the gap with large companies for basic process tools—websites, online sales, accounting, and human-resources systems—but continue to lag on customer-relationship management, enterprise-resource planning, and advanced technologies. Micro firms with five to nine employees look broadly similar to small firms for e-commerce, online marketing, and cloud computing, yet fall behind on systems and advanced tools.
The difference is not simply that advanced technologies are newer. They often require more than purchasing a product. They can change workflows, demand new skills, require reliable data, and impose a larger change-management burden. A mature cloud service may be adopted as a relatively bounded improvement. An advanced automation or AI project may require the firm to redesign processes and decide how the technology fits its specific operations. The theoretical non-rivality of the resulting software does not remove those initial and complementary costs.
The survey evidence reflects that size gradient. In the UK Innovation Survey, about 17.5% of SMEs and 24.5% of large businesses invested in software in 2020–2022, while the corresponding hardware investment rates were much closer. The figures are not a direct test of the U.S. Census mechanism: they use a different country, survey, period, and definition of software investment. They do, however, illustrate a relevant practical condition. Software adoption can be more size-sensitive than hardware investment when implementation, management, and uncertainty are part of the investment decision.
That condition maps onto the threshold in the firm-scope model. The model represents the fixed and variable economics of adoption in a compact way. The UK report supplies concrete categories of friction that can raise the effective threshold: implementation and maintenance costs, subscription costs, skills gaps, uncertainty about returns, and the risk that a badly integrated system disrupts existing operations. A small firm cannot necessarily absorb a failed project, even if a successful system could eventually be reused across several sites.
Technology maturity also affects the evidence available to a prospective adopter. A mature tool has more examples, established vendors, and familiar implementation practices. An emerging tool may have a larger long-run potential but less reliable information about its use cases, security, integration, and returns. This makes “wait and see” a rational response under uncertainty, not merely resistance to change. Firms may wait for peer validation because another firm's experience reduces the risk of making an irreversible or disruptive investment.
The same capability can therefore have different value in different firms. A multi-establishment manufacturer with common production routines may be able to reuse an automation or planning system across sites.