More than 80% of buy-side portfolio managers and analysts are now supported by their firm’s programmatic research infrastructure, as are about 40% of risk and compliance teams, according to the results of a new study from Crisil Coalition Greenwich, in partnership with Bloomberg.
Currently, about a quarter of participants in the study say their programmatic research infrastructure is fully centralized, while roughly half support programmatic research in a hybrid manner, in which some tools and processes are centralized, while others are not.
Many firms recognize the value of sharing programmatic research, but in practice, reuse is often “sporadic at best.”
According to Audrey Costabile, Senior Analyst in Market Structure & Technology at Crisil Coalition Greenwich and author of ‘Buy-Side Programmatic Research Infrastructure and Tooling’, the single biggest obstacle are “structural silos that are present in many firms”.
“The way firms are designed often discourages sharing,” she said.
For instance, she said, the most common barrier cited is that different teams have fundamentally different use cases. As one quantitative researcher explained, there’s lots of reuse within his Global Equity team, but between Global Equity and Private Equity, “there are just different business models and a lot of things in one team that aren’t relative to the other.”
This leads to tools being built for “one purpose and one team only,” as another respondent noted, according to Costabile.
“The tool is so optimized for one team’s workflow that it’s practically useless to another without a major refactoring effort, which no one is incentivized to do,” she told Traders Magazine.
A human element may also be part of the problem, she noted.
“There’s a natural tendency for teams to trust their own work more than anyone else’s,” she said.
According to Costabile, one analyst mentioned that using data from another team is tough because it’s “tough to trust.”
“In a competitive environment, sharing a promising model or idea before it’s fully proven can feel like a risk. Meanwhile, the simplest but common barrier is a lack of communication,” she said.
With buy-side decision-makers split on whether current infrastructure is future-proof, “foundational plumbing could buckle first because current methods of moving and preparing data are often manual and aren’t built for the scale and speed- AI requires,” Costabile stressed.
“An analyst today might spend hours “cobbling together worksheets and various other data feeds”,” she said.
When an AI model needs to access and process that same data from multiple sources in real-time across the entire firm, this manual “plumbing” may fail, she said, adding that the report emphasis the importance of “shared data pipelines”.
Humans may also “break” because of the increased demand for enhanced skillsets to support research, according to Costabile.
“We mention a lack of expertise as an underlying factor that breaks down collaboration. There’s definitely a skills gap for some technologies that’s expensive to remediate,” she said.
She added, lastly, governance and compliance could become problematic because a model today might be restricted to a specific team’s data on a “limited share,” as one respondent described.
When a new, firm-wide AI initiative needs access to data from the equity, fixed income, and risk teams simultaneously, these siloed permissions could the project to a halt, Costabile said, adding that
many participants are turning to vendors to solve these issues rather than trying to tackle them internally.
As third-party vendors play a larger role in core research infrastructure, firms are outsourcing execution rather than research strategy so vendor relationships are a tool in this sense to deliver whatever vision professionals at a firm have in mind, according to Costabile.
Many firms have vendor competency teams that oversee the relationship, procurement, monitor usage and so on for e.g., firm-wide tools like Bloomberg, she said.
“This team can ensure the vendor solution(s) is delivering value,” she said.
“Finally, most participants tell us they use a hybrid programmatic research infrastructure. They typically aren’t 100% reliant on vendors for a few reasons: “last mile” engineering is usually is very firm-specific and requires a great deal of flexibility,” she added.
According to Costabile, often, firms also mix a number of vendor solutions and data sources that a single platform can’t manage – there is likely a need for a cohesive infrastructure.
In this case, vendors are partners but ultimately there are important internal infrastructures supporting operations, she said.
“Essentially all the pieces have to fit together. On the other hand, smaller firms with fewer resources may benefit from relying more heavily on vendors since they can help with compliance, audit, tooling and more – leaving the firm with resources to focus on analysis and investment outcomes,” she concluded.
