The Science of Artificial Intelligence and Alternative Data Isn’t Fiction Anymore
The financial industry is awash in massive amounts of data. How that data gets harnessed, analyzed and used will increasingly not be up to traders and asset managers, but rather machines that, eventually, will operate faster, smarter, and with a razor-sharp predictive view of what’s to come in the markets.
Science fiction this isn’t
The rise of artificial intelligence, a concept that has been around for more than 50 years, has gained tremendous momentum recently due to breakthroughs in machine- and deep-learning, cognitive analytics and robotic process automation technologies. This latest AI wave is unlikely to fade, thanks to affordable, large computing power, coupled with access to unprecedented amounts of data, including new, alternative data sources.
Given the financial world’s addiction to data and information, it’s no wonder that AI technology holds so much promise for industry participants. It can capture structured and unstructured data from varied sources and leverage that information to improve business decisions. It can predict, hypothesise and act with other systems autonomously. The next frontier for AI will be to deliver pertinent reasoning and explanation for an unforeseen event.
Capital markets players are starting to make big investments in the powerful combination of artificial intelligence and alternative data. Their intention: to generate significant savings and alpha ahead of rivals and FinTech start-ups.
While still in relatively early stages—AI, for example, can’t yet adapt to every scenario and all situations—capital markets firms are making significant investments in AI technologies and alternative data, which reflect their relevance and potential. In 2017, Opimas expects spending on AI-related technologies to exceed US$1.5 billion and, by 2021, to reach US$2.8 billion, representing an increase of 75%. This does not include M&A activity and investment in start-ups made by financial institutions.
At the same time, trading and asset management are being fundamentally changed by an explosion of information that is reshaping institutional operations and creating a new hierarchy of winners. To stay competitive, traders and portfolio managers will increasingly need to incorporate alternative data that stretch well beyond the traditional market intelligence that has been the mainstay of investing.
These alternative data come from a bewildering array of sources, including satellite and drone imagery, GPS tracking for cars, trains, and mobile phones, transactional data for credit cards and other payments, sentiment analysis for social media and news feeds, and so on. Frequently, the new data was not designed for investing, but rather for marketing, agricultural and industrial production, security and other purposes.
Opimas estimates that buy- and sell-side investments in the race to master and employ the plethora of data will exceed $US7 billion by 2020. While potential returns are currently impossible to pin down, the current 21% annual growth in spending in this space implies that managers believe the alpha generated will at least cover their investments initially, and over time could produce attractive returns.
It’s not surprising that the industry has high hopes for AI and emerging alternative data sources, given the unsatisfactory levels of operational efficiency at most capital markets firms and their desire to exploit the vast quantities of data that they generate. It also appears that the investments they have made in recent years in big data technology have not been wasted.
The asset management industry will be the most impacted with a reduction of 90,000 jobs, as AI will intensify clients’ disenchantment with traditional asset managers and lead them increasingly to cheaper, automated strategies. On the other end, we expect close to 30,000 new positions to be created at technology and data providers to satisfy the new and growing demands generated by AI.
Convergence of Trading Models
To date, AI has allowed financial firms’ software to make use of unstructured data, thanks to image recognition, natural language processing, etc. On the trading front, for example, there are already numerous adoptions of AI technology, most of them tied to cognitive analytics and machine learning, the details of which are closely guarded. Opimas is aware that their overall objectives are to leverage different types of quantitative (e.g., market data, annual reports) and qualitative (e.g., social media, news feeds, etc.) inputs to drive the convergence of systematic trading and quantitative fundamental models.
Figure 1. AI and Alternative Data support the emergence of new trading model
The explosion of alterative data will require hedge funds and other asset managers to make large investments to acquire the necessary skills and infrastructure to leverage these sources of information. We expect that alternative data will contribute significantly to a further shrinkage in the hedge fund population, as firms unable to exploit the information needed to compete effectively in the new world of intelligent investing will fall behind. The sell side will be challenged, too.
This sea change in asset management creates opportunities for providers of the underlying data, but also presents some challenges. Buy-side firms will need to develop a wide range of skills including data management to unearth and apply hundreds of data sources covering thousands of data sets; domain expertise to interpret and contextualise the new data; data science to create quantitative trading models using advanced statistics and artificial intelligence; and information technology to create architectures designed to deploy the investing models based on heterogeneous data. Sell-side institutions could benefit by providing the necessary infrastructure to their buy-side clients, while broker-dealers that are part of large universal banks are well-positioned to package and resell some of the vast quantities of data to which they have access internally. Market data vendors are positioned to act as aggregators of fragmented data sources and to provide services that directly tie the alternative data to tradable instruments. Exchanges are likely to see muted growth in their more advanced market-data offerings. Vulnerable areas will include ultra-low-latency data feeds and depth-of-book products, as clients’ earlier focus on quantitative trading shifts away from high-frequency strategies to those that are rooted in alternative data.
Typically, these raw data sources are not in a format that lends itself directly to investing, and require considerable processing to extract data that can be fed into quantitative models. While some leading investment firms may invest the resources to perform these analytics themselves, many firms have appeared on the market who run their analytics on the raw data and sell the resulting data sets.
AI at Center of Paradigm Shift
AI is likely to become the foundation on which quantitative investing models are built, fueled by swarms of emerging alternative data. Managing this flow of information and converting it into investment strategies will be one of the biggest challenges facing the asset management industry in the coming years. Gathering the alternative data will not be easy either, because the market is highly fragmented, with a rapidly increasing number of sources emerging.
The horizon for a potentially profitable use of alternative data seems wide open for the foreseeable future, while the traditional quantitative strategies have lost value as too many market participants pursued similar methods. Not only are there hundreds of providers of alternative data already operating, mostly occupying very small niches, there is an explosion in the number of data sources underway, as well as in the utility of the information they can provide.
Figure 2.Shifting Model in the Value of Data
While satellite imagery gets a lot of attention, there is a plethora of other data sources that is growing rapidly. The major task for investment firms will be to seek out new, innovative sources of data and to integrate them into their trading strategies. This process will be fraught with difficulty. While the skills acquired creating quantitative trading strategies with traditional data sources certainly will apply in this new world of information, there are major differences. Alternative data, by its very nature, is fragmented rather than consolidated; periodic rather than continuous, and unstructured rather than normalised. Finding and leveraging alternative data sources requires new skills and significant investments that will exceed the resources of many asset managers.
Currently, it is impossible to pin down how much in excess returns, or alpha, might come from an intelligent use of alternative data sets. The only real answer is an unsatisfying “it depends.” However, we gained insight into how much alpha certain firms believe they are generating from an intelligent use of alternative data by examining a leading quantitative hedge fund with whom we have worked. The hedge fund spends more than US$50 million annually on sourcing, analysing, and implementing alternative data strategies. This implies the fund managers believe they must generate at least about 2% in alpha annually simply to cover the associated costs. The likely actual number may be much higher. A conservative estimate would be that alpha generation of at least 4-5% is possible by effectively harnessing alternative data sources. Similarly situated firms may anticipate similar results.
The powerful combination of AI and alternative data will enable more sophisticated trading strategies with potentially higher returns, but it requires firms to make significant investments and tap into a limited pool of expertise. This will restrict those who can really play ball, and Opimas believes the emerging new business model will be the death knell for many traditional asset management firms as they will not be able to adapt to this radically changing landscape. We expect a consolidation of the hedge fund industry to accelerate in the coming years, as successful players gain market share from firms unable to adapt to the technological revolution that AI and alternative data bring to bear on the capital markets.