Cultivating a machine learning culture at asset management firms
Machine learning culture thrives in environments that encourage cooperation, as well as data, technology and model sharing
MACHINE learning has vast potential for asset management firms. Deploying systems that use data-driven technologies can streamline buy-side operational processes, unearth valuable insights and possibly improve portfolio strategies and performance. But only firms that embrace a machine learning culture—and are open to applying machine learning techniques creatively and broadly—have the best chance of reaping the benefits.
To build a culture that extracts the most value from data and machine learning, it’s important for buy-side firms to commit senior management to this effort to foster firm-wide collaboration and investment in data-driven technology. Firms should also approach this cultural transformation as they would any strategic shift across the organization, allocating capital and rethinking the personnel and skills necessary to make it happen.
The data revolution lights the spark
Stepping back, the burgeoning data revolution and the proliferation of artificial intelligence (AI) continue to reshape industries and sectors, including asset management.
As a subset of AI, machine learning involves systems using data to learn, improve and construct analytical models with significantly less programming effort or even the ability to solve previously intractable problems. As the amount of data has expanded exponentially over the past two decades, so, too, has the capacity for data gathering, storage and processing.
Alongside this growth, evolving computing technology has empowered market participants to scale out collecting, storing, cleansing and analyzing ever-greater—as well as more diverse and granular—amounts of structured and unstructured data from increasingly varied sources at greater speed.
The mainstreaming of data science, coupled with computing advances in AI, have allowed asset managers to successfully scrutinise large and complex quantities of data for patterns. They hope to harness this technology in ways that boost returns, enhance strategies, lower costs and improve processes and operations.
Ultimately, investors anticipate data and powerful analytics will continue to improve many front- and middle-office functions as they reshape the investment management landscape. These include portfolio construction and rebalancing, risk management, compliance and trade-cost analysis, among other benefits.
The spectrum of data-driven technology use
In the asset management space, the levels of data-driven technology sophistication, adaptation and application vary among firms with different investing objectives. Mutual funds and larger, fundamental-type asset managers, who are predominantly conservative investors in mostly secure assets, take a relatively lower-tech, human-driven approach to asset selection. Mutual funds, which mostly design portfolios that employ strategies based on classic factor models, generally use the necessary technology required for that process.
Meanwhile, hedge funds tend to come in two varieties, both of which lean on data-driven technologies more heavily. Many hedge fund investors use strategies similar to those of mutual funds, but which can take on more risk, apply leverage and are more interested in deploying technology to find an advantage. Other more advanced hedge funds adopt exotic and sophisticated investment strategies and embrace data-driven tools and analytics as a core alpha-generating tool.
Data culture at buy-side firms shifted, mostly over the last 10 years, toward more organizations adopting data-driven approaches to their strategies and models. Technologically advanced, well- capitalized hedge funds demonstrated early data use and expertise and today make these technologies a part of their daily portfolio reallocation and risk strategy.
Fund managers at some larger firms and mutual funds followed in using data and machine learning in their front and middle offices. And large asset managers, mutual funds and core fundamental shops have cautiously followed into the data space, mainly concerned with learning where and how to alter processes and group culture to be more collaborative around data that currently tend to be siloed.
All these efforts are important because the benefits are legion. Though many traders and portfolio managers aspire to find consistent ways data can help them develop alpha-generating strategies, firms more broadly look for this technology to improve many front- and middle-office functions. These include:
- portfolio construction and rebalancing
- risk management compliance and regulatory monitoring and reporting
- trade-cost analysis
- asset allocation
- irregular transaction identification
- clerical and other task automation
- general collaboration across groups
Start at the top
How do asset management firms cultivate a machine learning culture? They can begin through securing buy-in on the importance of machine learning from senior management. Gaining leadership’s commitment makes it easier to establish and spread a more data-centric culture across the firm so that it becomes the core part of messaging, DNA and planning.
Furthermore, this machine learning culture thrives in environments that encourage cooperation, as well as data, technology and model sharing. For example, teams covering different asset classes can share data techniques to solve similar but distinct problems or communicate valuable information more easily across silos.
Asset management firms that have a machine learning culture tend to be flexible, as well as particularly open to collaboration, new ideas and novel ways of thinking. They develop adaptable decision-making processes around the types of models they use. And they consider investments in machine learning applications at all levels of the firm, not just in the front office.
Regard the question as a strategic one
Organizations can approach the matter of developing a machine learning culture much as they would any kind of strategic planning, such as buying a new building or expanding into new asset classes. The firm can diagnose the strategic imperative it’s trying to address and determine whether it can apply data-driven solutions. It then can determine how much capital to allocate to this effort.
To do this properly requires the right personnel. Asset managers can start by hiring a senior-level data or machine-learning expert with experience transforming firms in the financial industry who understands the organization, its strategy and culture. This expert should develop multiple options for a multi-phase plan to evolve the organization in a way that is consistent with culture and objectives.
At the next stage, senior management can explore what it would mean to make staff “machine-enabled.” Future talent will require skillsets that are more data-proficient. As many across Wall Street worry about automation and job loss, organizations should be sowing the seeds now to ensure staff are more fluent in data-driven technologies, whether they work in the front or middle office, at a fundamental or more quantitative firm.
Workforces will need to evolve as technology evolves. So, as part of a machine learning culture, firms should consider how to make sure their employees have the right skillset to move forward with the changes in the industry.
The data explosion holds promise for any firm that can process complex and disparate information sources quickly, efficiently and creatively. Asset managers who explore data-driven technologies’ potential fully put themselves in a strong position to stand apart from their competition, improve performance and attract new investors.
Naz Quadri is global head of machine learning and alternative data at Bloomberg
27 Nov 2019