The mechanisms of hypergrowth
Why do some companies quickly reach a capitalization of 1 billion dollars or euros, while others contemplate this emblematic threshold without ever crossing it? What makes the difference between a potential unicorn and a simple scaleup among others? This question worries enough people that eminent authors have looked into it.
Out of 1000 scaleups, how many become unicorns ?
Aileen Lee coined the word Unicorn in 2013. The term appealed to her because it unambiguously evokes something rare. Indeed, few startups reach a valuation equivalent to $1 billion less than 10 years after their creation. Even when focusing on projects centered on technological innovations, the success rate remains extremely low. It is even lower if we only consider independent companies, i.e. not backed by large groups, and not listed on the stock exchange. In your opinion, out of 1,000 startups of this type, how many will one day achieve this? The statistic was measured on a large panel of American structures financed by venture capital firms. From around 1 in 1,000 in 2013, to 1 percent 10 years later, in 2023. While innovation managers have made significant progress, they are still far from reaching a level of maturity. A factor of 10 in 10 years is certainly very respectable, but the most ambitious investors are still faced with an enormous level of uncertainty.
This article is reserved for subscribers. Already a subscriber ? Just log in. Or subscribe for 1,80 € (90 % of the article left to read - available in English and French).
How do investors in innovation choose their targets ?
The daughter of Chinese immigrants from New Jersey, Aileen Lee attended Harvard before becoming a leading financial analyst in the world of technology entrepreneurship. Today, she is both a theoretician and a practitioner. She founded and runs a venture capital firm, while mastering the modeling of the companies in which she invests. Although she does not strictly belong to the academic world, she strongly influences the work of many researchers.
Sebastian Krakowski is one of them. The title of one of his papers immediately catches the eye. It begins with Using deep Learning to Find the Next Unicorn. Basically, how can we harness the power of artificial intelligence to find the next unicorn? Or, if we put ourselves in the shoes of a scaleup manager, how can AI help me find the right strategy to take the next step in my development? Without Aileen Lee, this study would never have existed as such. Thanks to her.
Sebastian Krakowski is a young researcher who values ​​the work of his elders. He lists previous publications on methods for detecting future unicorns, before drawing his own conclusions. He worked with the global venture capital fund EQT Ventures, of Swedish origin, whose investment target selection tool exploits the potential of deep learning. It should be noted that their rate of discovery of future unicorns is approximately 10 times higher than the market average, namely around 10%. This mixed team, composed of both public and private actors, listed the practices of their peers around the world, other researchers and practitioners who are keen to detect high-potential start-ups as early as possible. In particular, they highlighted two sets of variables that are particularly interesting for anyone interested in the effectiveness of financing innovation, whether novice or experienced.
How to measure the success of a scaleup ?
The first set concerns success criteria, i.e. the conditions that allow us to confirm that a company has reached its Grail. The second covers success factors, i.e. upstream signals, precursors of upcoming hyper-growth. Let us summarize their results.
The success criteria can be grouped into categories, starting with the most important: the success encountered in terms of financing. Were the announced objectives met, at what speed, were they exceeded and confirmed by other significant fundraisings not initially anticipated, by the support of venture capital funds, by an IPO, by a buyout by a blue chip? What total amount of financing was obtained? These criteria are considered much more important than profitability or growth by specialists, even if one can easily imagine that both encourage investors to finance the initiative.
The second grouping of criteria, again in order of importance, concerns the sustainability and valuation of the company. Is it alive several years after its creation, has its workforce increased at an appropriate rate, is its valuation increasing, including during the latest financing operations?
Finally, the third and final theme is reputation: do analyses by recognized experts recommend investing in this company, has it won competitions or entrepreneurial challenges, does it enjoy enviable positions in the most followed rankings?
Is the impact of a scaleup correlated to its performance ?
Curiously, according to Sebastian Krakowski and his peers, no team of researchers or practitioners who have published on the subject would mention any impact on the ecosystem, a remarkable environmental, social or societal virtue. Yet even the most cynical of managers are sensitive to this! The reasoning is as follows: if the playing field on which the company operates is damaged, if its potential targets are not very solvent, if household morale is at its lowest, savings take precedence over consumption; unless you offer savings products, precisely, it is extremely difficult to develop in a market lacking dynamism, including by having a disruptive innovation. And there obviously cannot only be banks, asset or wealth managers to the exclusion of any other economic activity, otherwise what would these financial institutions invest in? The typical example is a unicorn positioned in the wood sector that would deforest faster than any of its competitors in history; it would end up drying up its own raw material. Of course, we can consider that the maintenance of a supporting ecosystem is the responsibility of the state or other companies – in short, “it’s not my problem, others take care of it” – or that there are always very attractive market segments and places in the world, for example luxury in Asia.
But it is clear that the most structuring levers of our societies in the 21st century are operated by hyper-growing private actors. If unicorns do not integrate a responsible agenda into their DNA, our humanity withers in the short term, then fades and collapses in the long term. Because GAMAMs, NATUs and other BATXs were first unicorns, and unicorns are future GAMAMs, NATUs and BATXs.
Should this aspect be considered a sub-topic of reputation, the third group of criteria mentioned above? This seems too restrictive. In reality, there is at least one study published on responsible hyper-growth. It remains to bring this movement to maturity. It is a question of transforming a vague concern into a family of useful and justified criteria, without falling into the trap of hyper-normalization.
How to anticipate hypergrowth ?
Let's go back to the conclusions of the paper by Sebastian Krakowski and his peers: Using deep Learning to Find the Next Unicorn. After identifying success criteria, which are used to confirm on the day that success is on the horizon, they identify success factors, these precursor signs that show that we are on the right trajectory. As you might expect, the goal here is to establish links, correlations between factors and success criteria. If such factors are observed upstream, then such success criteria are validated downstream, by scrutinizing a large number of sources, documents, studies, analyses, testimonies, in formats as varied as tables, texts, graphs, time series, images, videos or audio documents. Here too, let's group factors into 3 categories: people, assets and models.
Investors like successful individuals. They reassure them. This concerns founders as well as shareholders, mobilized teams, partners, members of formal and informal networks, right down to customer profiles and behaviors – the famous personae. On average, people-related factors account for about a third of the overall score if we summarize existing models in universities and among practitioners who have published on the subject.
The second category covers the state of assets in the broad sense, analyzing the financing history, the intellectual property portfolio, the efficiency of R&D, mergers and acquisitions events, financial performance (when data is available), and finally external data such as third-party ratings. It accounts for just under a quarter of the overall score.
Finally, the model category integrates the notions of product and service, catalog, positioning, competition, visibility, communication, processes and operational efficiency. It is predominant with 45% of the score.
Does the pedigree of key people attract investors ?
The weight of people-related factors seems reassuring. It would echo our confidence in humanity, no more and no less than in financial or operational considerations. In practice, it makes the breakthrough of outsiders more difficult. If the success of a financing round must go through the integration into the shareholders of a serial entrepreneur who has proven himself, this means that innovation is controlled by a small club of very influential people, with a major risk: locking oneself into the reproduction of what has already worked, to the detriment of truly new approaches.
Are scale-uppers optimizing their assets ?
The more modest weight of the assets category is explained by the difficulty in collecting reliable data, especially if we adopt the point of view of an investor looking for a target, having access only to exogenous information and not to the income statements and balance sheets of the structures scrutinized by their robots or analysts. In reality, few managers themselves are able to identify the non-financial assets of their structure, therefore not appearing on the balance sheet, and structurally under-exploit their potential. This could be, for example, the result of an abandoned prototype, a piece of software used exclusively internally but which would benefit from being promoted to customers, or an under-used electronic showcase. This phenomenon is also common at the level of liabilities, these unprofitable investments, these expenses without added value, or these stubbornness in inconclusive lines of investigation. It is virtuous to arbitrate excess time spent raising funds or participating in seminars and conventions, in favor of time spent evaluating one's own assets.
Are scale-uppers optimizing their business model ?
The importance of the factors in the model category is explained by the relative availability of data both outside and inside the company, with prices and designations of products and services being by definition visible, as are communication actions or competitive fields and to a certain extent the processes exposed to customers. Some generative AIs used to glean such information are also surprisingly efficient, provided that the right sequence of prompts is set with the right level of precision.
Can generative AI help optimize a business model ?
More generally, instead of favoring specific deep learning chains like the authors of the article in question, is it possible to design precise “prompts” that make a generative AI capable of selecting startups or scaleups with unicorn potential? Better still, is it possible to determine which actions, strategic decisions, and possible pivots would bring these companies closer to a more promising model, or even the most promising possible?
The first advantage of such an approach is linked to the greater openness of the model. For example, instead of filtering companies that are a priori incompatible with the displayed preferences of an investment fund looking for new targets, in an understandable concern to optimize the size of the source databases, all the knowledge of generative AI is mobilized.
The tools to be set up and the infrastructure costs are reduced to their simplest expression: the license of a generative AI favoring the authenticity of the result, duly traced thanks to a pointing to the sources used and a transparent, step-by-step explanation of the progress of the work carried out. Perplexity.ai for example is perfectly suited to this task.
Finally, it makes it easier to test the handling of qualitative subjects, which by definition lend themselves well to a text-based approach. For example, it becomes possible to take into account a central phenomenon in assessing the difference between a future unicorn and an eternal scaleup: the advent of a disruptive innovation, with a before and after for an entire ecosystem. The life of an innovation project is not continuous; it is made up of discrete events (in the “opposite of continuous” sense) that are more or less predictable.
How to integrate qualitative criteria in the evaluation of a scale-up ?
All of this converges towards a lack, a huge gap to be filled: unicorn detection models are far too quantitative or systematic, seeking above all objective data to the detriment of intuition and the integration of new knowledge.
Even the knighting of key people is a mechanical process: such and such a person must have already succeeded in order to be able to succeed again. These models are lacking in innovative approaches that take into account qualitative elements, reflections of the true creativity of the structure. However, this is where the bulk of sustainable differentiation occurs. This is where the potential for progress lies in the ratio of the number of unicorns to the number of innovative company creations, to go, why not, from 1% to 1 in 10. You still need to be well-versed in agile strategic analysis methods adapted to hyper-growth companies, as are the economic leaders of our time. See on this subject “Elon Musk’s Holidays”, which can be found in another article. Making strategy in such a context does not begin with the handling of figures or statistics, but by identifying qualitative potentials using methods that are both modern and proven. In this regard, strategic innovation management is a booming discipline, whose key challenge is improving ROI. After decades of blind trial and error, against a backdrop of accelerating breakthroughs in disruptive technologies, innovation is reaching its maturation plateau.
Post-hypergrowth : what are the prospects ?
Let’s project ourselves beyond this level, typically in 5 to 10 years. AIs full of data detect holes in the racket, these innovations that we are missing without even realizing it. They sort and order these opportunities, virtually matching them to hybrid teams, combining humans and machines, the most capable of carrying them out. They evaluate their chances of success, which they multiply by an impact score. And they start again. They are constantly recomposing a portfolio of more or less advanced initiatives, but which share a common characteristic that we can anticipate today: their development cycle is constantly shortening.