This chapter pushes further in our investigation of the nature of scale-ups by introducing the following eight propositions about scale-ups: (1) a scale up is a concept born of practitioners, not academics; (2) scale-ups are not just in the IT sector, but may be enabled by IT; (3) a scale-up is a qualitative concept from a ‘stages-of-growth’ model; (4) scaling up involves structural transformation; (5) a scale-up does not exist anywhere in a pure form; (6) scale-ups differ by degree, not by kind; (7) many scale-ups are exceptions; and (8) there may never be a standardized empirical definition of scale-up. Each of these eight propositions was either not clear, or was misunderstood, in some previous work.

4.1 A Scale-up Is a Concept Born of Practitioners, Not Academics

Scaling was a ‘hot topic’ in the practitioner literature for a long time, while being conspicuously absent in the academic literature (Shepherd and Patzelt, 2022; Jansen et al., 2023). The concept of scale-up is closely linked to the Silicon Valley venture capital scene. Scaling up concepts have long been analyzed by authors such as Steve Blank (2013), Furr and Ahlstrom (2011), and Eric Ries (2011) from the ‘Lean Startup’ movement, as well as Reid Hoffman, co-founder of PayPal and LinkedIn (Sullivan, 2016; Hoffman and Yeh, 2018; see also Kuratko et al., 2020), and Peter Thiel, co-founder of PayPal and Palantir, and the first outside investor in Facebook (Masters and Thiel, 2014), and also bloggers such as Paul Graham, co-founder of the startup accelerator Y-combinator (e.g. Graham, 2013). These authors went far beyond the confines of their ‘day jobs’ (i.e. investing and supporting startups) to develop concepts, share ideas (while sometimes respectfully disagreeing on some points),Footnote 1 and grow a community of rigorous thinking about the science behind new venture performance. For example, Graham’s 2013 blog post benefitted from interactions with peers including Sam Altman (CEO of OpenAI) and Patrick Collison (co-founder and CEO of Stripe).

Scale-ups have also received attention from policymakers (e.g., OECD, 2021), with a number of policy reports and ‘grey literature’ focusing on the phenomenon of scale-ups (e.g. Coutu, 2014; Hellmann and Kavadias, 2016; Duruflé et al., 2017; Reypens et al., 2020; Vandresse et al., 2023). Meanwhile, it seems that academics are playing catch-up; still trying to determine what exactly a scale-up is and how to define it. Academics started exploring the concept of scaling-up starting with publications such as Sutton and Rao (2016) and DeSantola and Gulati (2017), and some recent journal special issues at the Journal of Management Studies (Jansen et al., 2023), Journal of Business Venturing (Autio et al., 2021) and Journal of World Business.Footnote 2 These early academic publications did not give specific details regarding the definition and identification of scale-ups, however: Sutton and Rao (2016) (as well as follow-up work by Shepherd and Patzelt, 2022) focus their discussion on the important topic of how to scale-up excellent practices throughout a growing organization, although they do not define a scale-up and some of their examples even seem to refer to firms that have not grown in a long time.Footnote 3 Meanwhile, DeSantola and Gulati (2017) discuss the organizational transformations in growing firms at a relatively abstract theoretical level.

4.2 Scale-ups Are Not Just in the IT Sector, But May Be Enabled by IT

Scale-ups have repeatedly been linked to digitalization and the IT sector (Adner et al., 2019). “Software has a natural affinity with blitzscaling, because the marginal costs of serving any size market are virtually zero” (Sullivan, 2016, p. 46). IT is a GPT (General Purpose Technology) that permeates all sectors and lowers the costs of scaling by reducing the costs of distribution, storage, processing and replication of data for a broad set of tasks in all sectors. Unlike physical assets, digital assets such as databases can be “reused, repackaged, and resold ad infinitum” (Adner et al., 2019, p. 257), such that data continue to provide value for a firm even if it has already been applied in different contexts. This can help firms to overcome the usual limits to growth (i.e. Penrose effects, Piaskowska et al., 2021, p. 5). The development of IT has also enabled the phenomenon of viral marketing, which is enabled by IT and social media, although the benefits of viral marketing are not confined to IT sectors but can benefit all sectors.

A first implication of digitalization is that distribution costs become close to zero for digital goods and services such as Software-as-a-Service (Adobe), music (Spotify) and video (Netflix) (Hoffman and Yeh, 2018). Distribution of goods via the internet gives a crushing advantage over distribution of physical goods via brick-and-mortar outlets, as evidenced by the disruption of Blockbuster video by Netflix.

A second implication is that products sold to customers over the internet (such as Software-as-a-Service) can be easily tweaked and updated (Nambisan et al., 2019). Software updates can be sent out to users that are implemented over an internet connection. An implication of this is that software products can be sold in an interim and incomplete condition, with the understanding that improvements will regularly be sent out to users to ensure that the security and performance of the software product are up-to-date. Software is clearly easier to update over an internet connection than hardware. This leads us to the 4th rule of blitz-scaling: “launch a product that embarrasses you” (Hoffman and Yeh, 2018, p. 206). Software companies can sell an embarrassingly bad version of their product as long as this can quickly be updated when the improved version is ready. Selling a premature and incomplete product could give software companies the edge they need, if they strive for a fast launch, pre-empting the competition, and dominating the market as they benefit from a virtuous cycle of positive demand-side network externalities (whereby new users prefer to join a network that already has a large number of established users, resulting in a situation whereby large increases in demand occur without much effort on the part of the producing firm (Mithani, 2023; Huang et al., 2017)).

Launching an embarrassingly bad product has been attempted by producers of physical products, with less success. One example would be Jawbone, a producer of Bluetooth headsets (Kuratko et al., 2020). Jawbone launched an incomplete product with the hope that imperfections could be fixed through software updates, although it transpired that the hardware itself also needed fixing, leading to customer refunds and a spectacular drop from its previous unicorn status valuation. A second example of a scale-up built around a physical product could be Tesla Motors (Hoffman and Yeh, 2018). Tesla sold the promise of Full Self Driving (FSD) mode that, when it became available, could be remotely retro-fitted (like a software update) on pre-existing Tesla vehicles (Niedermeyer, 2019). Such promises have failed to materialize, however, potentially leading to legal and financial liabilities.Footnote 4

Discussions of scaling seem to focus first on prototypical examples of software/IT firms, and then zoom out and concede that scaling-up also occurs in other sectors. While software-as-a-service seems particularly amenable to scaling-up (Nambisan et al., 2019), nevertheless any sector can have its scale-ups as long as it can accommodate firms with low marginal costs of scaling up their revenues (for example, via strategies of setting up their own platforms). Presumably if a scale-up exists in other sectors, it is likely that it is powered by IT capabilities in many cases, e.g. Uber in the taxi industry, Airbnb in the hotels industry, or the case of Copenhagen Seafood A/S discussed in Nielsen and Lund (2018). It would be surprising indeed if a scale-up did not even have its own website.

4.3 A Scale-up Is a Qualitative Concept from a “Stages of Growth” Model

A scale-up is a qualitative concept that emerged from the “stages of growth” school of thought—as discussed above in Chap. 3.2. Understanding scale-up as a growth stage is important and helps us realize that not all firms are well-positioned to launch into scaling if the background conditions and preparatory work have not been satisfactorily set up in place. Therefore, while scale-ups are younger on average, stages of growth models help us to understand that the very youngest firms are unlikely to become scale-ups because they are probably too young to have passed through the initial stages. A frequent problem among startups seems to be that they scale too early (Lee and Kim, 2023).

4.4 Scaling up Involves Structural Transformation

After a period of scaling up, a firm has a different structure than beforehand (Kimberly, 1976; Flamholtz and Randle, 2015), although such structural transformation is not necessarily required under the definition of HGFs. Growth by scaling up involves rapid increase in revenues but a negligible increase in costs, therefore the share of costs in total revenues is lower at the end than before scaling up. The capital intensity (assets/sales) of a scale-up is presumably very high initially (given the initial fixed costs of setting up the IT infrastructure and platform), although it is lower at the end of the period because the denominator (sales) has increased. Given their focus on platform strategies and ICT-powered business processes, scale-ups are presumably intensive in intangible assets such as software and IT infrastructure (Haskel and Westlake, 2017; De Ridder, 2023). Firms preparing to scale up will invest in the upfront costs of enlarging their corporate infrastructure, such as building bigger and better IT systems, and setting up centralized HRM systems (Von Krogh and Cusumano, 2001). The labor intensity (employees/sales) decreases over the course of the scale-up process because sales increase without requiring the addition of employees.

On the flipside, if a firm has the same structure (cost structure, capital intensity, labor share) after the growth event compared to before, then such a firm would not be a scale-up (although it may well satisfy the definitional requirements of an HGF).

Later on in this book, we suggest that an empirical definition of scale-ups (as opposed to regular-growth firms) could draw on the idea that scale-ups have a significantly different structure before and after the growth period.

4.5 A Scale-up Does Not Exist Anywhere in a Pure Form

A ‘pure form’ of scaling up would have the following characteristics. Scaling up is a stage in a startup’s life course whereby learning no longer occurs, because the firm exclusively leverages existing knowledge. Scaling up is a type of growth which incurs zero marginal costs, as firms strive to reduce the marginal costs of growth while frontloading the cost structure with higher fixed costs (De Ridder, 2023). Scale-ups, in pure form, would be entirely composed of scalable resources, and would not contain any non-scalable resources (Levinthal and Wu, 2010). Scaling up involves zero product refinements, because issues surrounding product development were fully addressed in the previous growth stage (c.f. the stages-of-growth model in chap. 3.2). Such a ‘pure form’ of a scale up presumably does not exist anywhere.

Firms can never, for long, increase output without incurring positive marginal costs or modifying their product design. Even digital firms are bundles of resources that include (to some extent) resources that are not fully scalable (Giustiziero et al., 2023; Levinthal and Wu, 2010). The idea of zero marginal costs is an attention-grabbing notion and an over-simplifying metaphor, but upon a moment’s reflection it cannot (for long) correspond to real-world growth.

In the context of internationalizing new ventures, “scaling internationally requires carefully synchronizing functional departments and value chain activities across geographies at a much larger scale. It involves coordination and mutual adjustment across domains of the organization to ensure coordination, foster collaboration, and reduce conflicts” (Jansen et al., 2023, p. 9). Therefore, scaling requires creating new functions, hiring new employees, and focusing attention on new problems.

Investments in fixed costs that might not have been worthwhile with a small customer base might suddenly become worthwhile when the customer base is ten times larger, such as strengthening the internet infrastructure for higher traffic, and streamlining the user experience in subtle almost-indistinguishable ways. Hence, growth of quantity may lead to increases in ‘fixed costs’, and ‘fixed costs’ may not be truly fixed at the start of the growth period.

Scaling is easier for digital firms compared to traditional manufacturing firms (e.g. because of low distribution costs; Kuratko et al., 2020), but even there it is not perfect. Digital firms have resource bundles that include non-digital resources (such as warehouses, inventory, and packing employees in the case of Amazon) and therefore cannot be costlessly scaled up. Giustiziero et al. (2023, p. 1396) continue along these lines:

For example, software and AI platforms need experienced engineers to develop, maintain, and improve them, marketers and salespersons to sell their outputs, customer service professionals to improve service quality, and managers to oversee and direct the enterprise. Often, physical resources such as factories, offices, and warehouses, and even hardware and telecommunication infrastructure to host and deliver digital products, are also required.

Even digital firms with scalable resources must somehow combine these with tangible harder-to-scale resources (Levinthal and Wu, 2010), such as for example allocating managerial attention to learning and the coordination of activities.

4.6 Scale-ups Differ by Degree, Not by Kind

A scale-up is an impossible ideal-type. It is an extreme and never-attained end of a continuum.Footnote 5 Instead of distinguishing between pure scale-ups and non-scale-ups, the task therefore is to distinguish between ‘almost-scale-ups’ and ‘not-so-much scale-ups’, and this latter distinction has a blurred boundary and necessarily leads to some arbitrariness in fixing a threshold between firms labelled as scale-ups and non-scale-ups.

Let us consider the simplistic case whereby a single variable (such as the marginal sales/cost ratio or perhaps a spike in marketing expenditure) can be used to distinguish between scale-up HGFs and non-scale-up-HGFs. As such, it would be theoretically interesting to see the empirical distribution of this, in the set of HGFs. For the conceptual diagrams below (Fig. 4.1), the relevant sample is HGFs only, and we compare scale-up HGFs with non-scale-up-HGFs. Figure 4.1 (left) below has no clear distinction between scale-ups and non-scale-up-HGFs. It is hard to make the case that scale-ups are a different category. In contrast, Fig. 4.1 (right) is a far more interesting case (in the sense of having theoretical clarity), showing a clear separation between scale-ups and non-scale-up-HGFs. In Fig. 4.1 (right), scale-ups appear to be a distinct phenomenon. Ideally, we would find a variable that can distinguish between scale-up HGFs and non-scale-up-HGFs in a way that shows clear bimodality (two peaks).Footnote 6 If such bimodality is not observed, however, then we would be in the case of Fig. 4.1 (left), where the distinction between scale-ups and non-scale-ups seems arbitrary. As we will see, Fig. 4.1 (left) fits the data much better than Fig. 4.1 (right).

Fig. 4.1
2 histograms plot frequency versus variations used to distinguish between scale-ups and non-scale-ups. The dotted line between the highest frequency exhibits the cutoff point for distinguishing between scale-ups and non-scale-up H G Fs.

Distinguishing between scale-ups and non-scale-up HGFs. Left: conceptual diagram of a unimodal distribution of sales/cost ratio for the set of HGFs. Right: conceptual diagram of a bimodal distribution of sales/cost ratio for the set of HGFs. [Color online]

Full size image

Figure 4.2 above analyzes our Swedish data (introduced in detail in Chapter 7). In line with previous research (e.g. Varga et al., 2023; Palmié et al., 2023), scale-ups are defined in Fig. 4.2 as the subset of HGFs that have relatively rapid growth of the sales/cost ratio over a three-year period (high sales growth, but negligible growth of costs).Footnote 7 This is an imperfect empirical definition of scale-ups (as discussed in Chapter 6), although it is illustrative and sufficient for our current purposes. The main message from Fig. 4.2 is that there is no natural dividing line to distinguish between a species of scale-ups compared to other HGFs. When measured in terms of growth of the sales/cost ratio, scale-ups are not a radically different breed of firm (i.e. not the case of Fig. 4.1 right), but that they are a category that is created by being beyond a somewhat arbitrarily-defined threshold (i.e. the case of Fig. 4.1 left).

Fig. 4.2
A histogram of density versus scgrowth, peaking at 7.5 for 0 scgrowth on the left. A line graph on the right illustrates the kernel density estimate, increasing, peaking at 0, then gradually decreasing.

Distribution of growth rates of the sales/cost ratio for Swedish firms. In this figure, sales/cost ratio is measured over a three-year period for the set of HGFs, and scale-ups (blue shading in Fig. 4.2 left) are distinguished from non-scale-up HGFs (orange shading in Fig. 4.2 left). Left: linear y axis. Right: logarithmic y axis. The dashed line corresponds to an arbitrary potential threshold of 0.2 (20%) for defining scale-ups. For a database description, see Chapter 7. [Color online]

Full size image

4.7 Too Many Exceptions

Scale-ups are, by nature, exceptional firms that find profitable markets by doing things differently. Hence, it is difficult to generalize about scale-ups because they are so unique. Practitioners pioneered the study of scale-ups from an angle of case studies and anecdotes (e.g. Hoffman and Yeh, 2018). If we try to apply a strict definition of a scale-up to large sample data, we might end up misclassifying a few scale-ups when comparing to what popular discourse considers to be “a real scale-up”.

Some authors claim that scale-ups should be profitable. But Uber is often called a scale-up, considered by some to even be a prototypical example of a scale-up (e.g. Pfotenhauer et al., 2022), and yet it was not profitable even up until its IPO in 2019. Kuratko et al. (2020) even go so far as to say that “most” scale-ups are not profitable.

Also, there are too many paradoxes. For example, a popular blogpost in the VC community suggests that scalers need to do things that don’t scale (Graham, 2013; see also Hoffman and Yeh, 2018), referred to as “throwaway work” (Hoffman and Yeh, 2018) that serves a temporary purpose as the firm moves down the long road towards ultimately ending up with highly-scalable business processes. Thus, the best way of production might involve capital-intensive high-fixed-cost intangibles-rich processes that take a long time to set up. But in the scale-up phase, you have to do things that can’t scale, like a CEO putting their personal mobile phone number as the customer complaints helpline (Hoffman and Yeh, 2018, p. 216), or Airbnb founders personally taking photos of the Airbnb rooms to help these rooms look more attractive on the internet (Hoffman and Yeh, 2018), or perhaps the case of Tesla aspiring towards solar power charging capabilities while actually using diesel-powered generators hidden round the back of charging stations to recharge the batteries of its cars (Niedermeyer, 2019).

More generally, Hoffman and Yeh (2018, p. 198) formulate a list of 9 counterintuitive rulesFootnote 8 of blitz-scaling, whereby scale-ups do the opposite of what we might expect from a scale-up.

Given that the origins of scale-ups are from anecdotal analysis from non-academics that do not particularly have the mindset of an applied statistician, a troupe of scale-ups will include lots of exceptions. These exceptions will be a headache as we try to formulate a generalizable definition of a scale-up that can be applied at scale to a broad range of datasets.

4.8 There May Never Be a Standardized Empirical Definition of Scale-up

While there is a fairly standardized definition of an HGF (Eurostat-OECD, 2007), a standardized definition of scale-up is more problematic. An empirical definition of a scale-up (which relates to firm growth as well as data on costs, marketing, and organizational structure) draws upon variables that may not always be available in a researcher’s dataset. We expect that researchers analyzing different datasets will not have access to the same variables as we have. Furthermore, unlike HGFs, the pure-form of a scale-up may not even exist in the real world (as discussed in Sect. 4.5).

Let us try to be as specific as possible. Scale-ups are a particular type of HGFs. Hence, there is a standardized definition of HGFs. Researchers might generally agree that scale-ups are a subset of HGFs, but they will probably disagree regarding exactly which subset of HGFs. Scale-up studies will probably agree on the first stage (the subset of HGFs), but find it more difficult to have identical operationalizations on the second stage (which subset of HGFs).

Some implications can be discussed. The probable lack of standardization in scale-up definitions makes it important for scholars to be cautious when generalizing across scale-up studies (especially for literature reviews and meta-analyses). Scale-up researchers should make it as clear as possible (e.g. in the abstract) how scale-ups are measured. Scale-up researchers should focus more closely on the previous scale-up papers that have a similar definition. Furthermore, the sensitivity of empirical findings to different scale-up definitions should be explored, where data allows.