Early Days For Buy-Side AI07.05.2019
The majority of asset managers are changing their investment process to use both structured and unstructured data but 40% have yet to see any impact from artificial intelligence on their investment process.
A new survey from Liquidnet, the institutional investor block trading network, said nearly two thirds, 64%, of asset managers now use both structured and unstructured datasets. The report, The Future of Asset Management, continued that almost three quarters, 73%, are using greater quantitative analysis in portfolio construction and 69% have increased their use of analytics to demonstrate execution performance.
Liquidnet spoke to 42 industry participants to discuss how they are changing their approach to data and technology between April and June this year. The asset managers who contributed are responsible for $11.4 ($1.25) trillion in assets and Liquidnet also interviewed data and technology providers.
The authors, Rebecca Healey, head of EMEA market structure and strategy and Charlotte Decuype, EMEA market structure and strategy at Liquidnet, noted the use of artificial intelligence to analyse unstructured data, such as software which systematically detects when CEOs “duck” questions on earnings calls to provide predictive analysis on future earnings. The report cited research by Barclays Capital which concluded that using this technique as part of sentiment analysis produced an annual return 13% above a benchmark index.
The head of trading at a global asset manager said in the study that their firm has a department just to procure data.
“There aren’t many that have the ability to consume and store that data – also the ability to cleanse it is important,” said the head of trading. “I hope the industry doesn’t find a way to make this easier as those with a technological advancement are in a better position to process it.”
Other examples of unstructured datasets being used include credit card transactions to track company earnings pre-estimates, company solvency statistics, satellite imagery to track shipping, sentiment analysis for consumer goods, or social media to track sentiment.
Firms are also scraping the internet and creating databases with less structured sources of information. A machine learning algorithm can then identify companies which have the same characteristics of previously successful investments.
“There is an increased appetite from asset managers based in Europe and Asia Pacific to incorporate data versus the US, which is surprising given the sporadic nature of unstructured datasets outside of the US,” said the report. “In addition, the cost of the individual dataset alongside the accuracy and cost of extracting value can be prohibitively high for many organisations.”
A managing partner and board member of a European asset manager said in the report that the biggest disruptors for the industry will be technology and data.
“Firms have to figure out how to be a good fintech firms, to better understand how to utilise and store data; and alternative pieces of data will be a big part of the asset management future,” added the managing partner.
Currently less than half, 40%, of fund managers have yet to see any impact from artificial intelligence on their investment process. The head of trading at a global asset manager said in the report that the firm uses some artificial intelligence techniques in software but it is “early days.”
“We are genuinely using and relying on AI in some of the algos we use,” added the head of trading. “We use clustering, but this is very narrowly focused as it is in a particular place in the business where it is fit for purpose. A highly mathematical quantitative model is not appropriate for everything in the business.”
Clustering techniques enable firms to group certain structured or unstructured datasets to help identify different data groups, establish what is noisy, irrelevant data to disregard and then by using the accurate subset of data, enable firms to ask the most relevant questions to extract the value.
Liquidnet predicted that access to richer and deeper datasets will shift from just verifying a portfolio manager’s investment idea towards the creation of ideas and firms also increasingly use quantitative analysis to understand manager behaviour.
“By being able to understand who is good at what, investment teams can figure out where the strong suits exist, invest more time and internal resources on these areas and outsource where there is opportunity to do this more effectively,” said the study.
For example, a multi-billion-dollar discretionary fund used systematic analysis to reviewed all research reports, emails, instant messages and trader notes to extract data signals and create a shadow portfolio. The report said that over a three-year period, using quant analysis together with the human-sourced content, the strategy outperformed the fund’s own P&L.
The head of digital product global asset manager said in the study that every organisation in asset management is using data and analytics.
“Now nothing is getting built right now that doesn’t have an API – it’s all about building cleaner and more assessable data,” added the executive. “Everyone on the street is using data, but my observation on how much you can use is how much you can invest into data and analytics. Rather than just building internally there is this new approach with partnerships.”
Liquidnet said the growing accessibility of unstructured datasets and new technology is leading to the digitalisation of the investment process. As a result sell-side research is switching from written analyst reports to deep industry expertise and the provision of proprietary datasets.
In May Liquidnet announced the acquisition of RSRCHXchange, which was launched in 2015 to distribute research from a variety of providers to asset managers through a centralized, cloud-based hub.
Brian Conroy, president of Liquidnet, told Markets Media at the time that the firm will launch a data offering to help asset managers capture more alpha following its acquisition of RSRCHXchange and OTAS in 2017.
OTAS was launched in 2011 to analyse market data and highlight actionable information for equities trading to fund managers in an easily digestible visual format. Software analyses changes from the normal pattern in data such as insider transactions, short interest, options and credit default spreads and automatically highlights the most relevant signals for stocks in their portfolio for review.
Conroy said the addition of RSRCHXchange allows Liquidnet to provide a new level of research and analytics more efficiently so the buy side can capture more alpha.
“We are running a pilot with large asset managers in the US over the next few months and will launch a cohesive offering that is driven by our members,” he added
Liquidnet’s report concluded that Investment in data and analytics is the first step towards essential digitalisation of the industry.
“The threat of artificial intelligence wiping out the asset managers is overdone – AI will not replace asset management, but those who invest in technology will undoubtably replace those who do not,” said the report.
Technology has enhanced capabilities of surveilling larger and more disparate data sets.
With Eugene Kanevsky, James Redbourn, and Joanna Wong, CLSA
AI/ML on the buy-side trading desk is a long-term program with short- and mid-term deliverables.
Aiden uses the computational power of deep reinforcement learning to improve trading results and insights.
BondDroid’s AI-generated prices are integrated directly into LTX’s pre-trade analytical tools.