The hardships faced by the buy-side these days are no secret—there’s the rise of passive investing, tightening margins and rising costs. And it’s also no secret that in order to find new sources of alpha, buy-side firms are increasingly looking to incorporate alternative datasets into their investment processes. Nevertheless, asset managers are still struggling with how to tap value from these products.

A year after data and analytics advisory firm Element22 released a white paper benchmarking the state of some of the largest asset managers in their journeys to use alt data and advanced analytics operationally, a second benchmark study by the company, sponsored by UBS Asset Management, has taken a broader look at the asset management space. Compared to last year’s sample size of 20 funds, 59 asset managers, representing more than 20% of global AuM, were surveyed. Forty—or 68%—of those were calculated to be in the first two stages of their journey—either the planning phase or trialing early proofs of concept. The remaining 19 were classified as being either in the middle of the road or as breaking new ground.

Our view is you need the data, but not all the data all the time for all the investments.

Suvrat Bansal, UBS Asset Management

What is clear from the study, says Predrag Dizdarevic, founding partner of Element22, is that many—if not three-quarters of the firms surveyed—were likely overestimating where they are on their alt-data journey. You can ask a person a qualitative question about how advanced they think they are, but what the quantitative questions showed—how many sources of alt data do they use, for how long, which other advanced technologies are they investing in, what types of machine learning do they use, and how much money are they investing in the program—is that they are likely not as far ahead as they might think.

“We had some questions on where they think they are in extracting the most value out of the use of these new technologies and the alt data sources, and their replies to those qualitative questions have been pretty aggressive,” Dizdarevic says.

When early-stage firms first dip their toes into the water, they can become overenthusiastic in thinking they’ll hit the gold mine fast, without realizing the amount of work that needs to be done to achieve sustainable value generation, he reasons. And despite smaller firms investing large percentages of their budgets into these areas, the numbers, in reality, probably aren’t as high as they’ll need to be.

“I think that realization will be coming later on,” Dizdarevic says.

The alt data space is a crowded and noisy one, and new providers pop up as fast as others fall. Suvrat Bansal, chief data officer and head of innovation at UBS Asset Management, says his firm had a huge advantage in being able to draw upon the lessons already learned in incorporating alt data within the bank’s Evidence Lab division.

“What we learned from that sell-side experience is these things have some level of timing to them. …We learned from it that no alternative dataset is useful all the time,” Bansal says. In other words, it’s a delicate balancing act where firms need to be able to adjust and try new things on the fly.

Recalling alt data’s countless hype cycles, he laughs. “It’s the ‘here-we-go-again’ [moment]. We know for sure in traditional asset management, it’s near impossible to use one source or, collectively, all sources. …There’s something within there, but our view is you need the data, but not all the data all the time for all the investments,” Bansal says.

There was a time when an individual dataset could equate to alpha, but those days are more or less gone, says Dizdarevic. Firms need to figure out how best to combine different datasets—say, satellite data, credit card information and mobile phone usage—to come up with unique insights. This isn’t necessarily a new phenomenon, but for firms that are still new to the alt data space, they may experience a period of disillusionment that, as he noted before, there isn’t an untapped individual gold mine dataset still waiting to be found.

Winds of Change

Abilities to re-think, change course, and decipher what will or won’t work are mission critical to the process for all firms, both Dizdarevic and Bansal say. For UBS, having to make adjustments came very early on. In one instance, the QED team ultimately had to weigh product generalization versus personalization when it came to natural language processing (NLP).

The general usefulness of NLP products could only be achieved if they addressed industry-common issues—researchers are overburdened, it’s hard to get insights, and transcripts are endless, Bansal says. But on a personalized level, the tasks at hand are how to engage people one-on-one through tailored use cases for idea generation, but the target has to be clearly defined.

Additionally, while some initial projects might seem fairly straight-forward at the start, firms need to be aware that there are usually additional investments that need to be made beyond the data, Bansal says.

It’s no longer your traditional systems. It’s no longer your traditional databases. It’s no longer your traditional coders. It’s a continuous blend of tech and portfolio management and research happening right in front of us.

Suvrat Bansal

“We underestimated, a bit, the capabilities on the infrastructure and data sides, which need to support these kinds of developments. You can today say, ‘I’m doing X, and I want to start producing two products on NLP,’ and that may require some local development with NLP models,” Bansal says. “But then it comes time to ask, ‘Are we ready with containers and cloud to run these loads?’ and you realize, wow—that’s not going to be two weeks!”

The winds of change in the asset management space are at the same time both intellectually exciting and tense. Beyond showing performance and returns, asset managers know they’ll have to personalize their services to align with individual clients, but that’s an incredibly difficult problem to scale, adds Bansal.

“It was one thing to enable one type of business…but that does not work for you anymore. If you want to drive unique capabilities, you need to combine those capabilities. You can’t just operate [by] optimizing each one. And I think that translates into the bigger problem,” Bansal says. “It’s no longer your traditional systems. It’s no longer your traditional databases. It’s no longer your traditional coders. It’s a continuous blend of tech and portfolio management and research happening right in front of us. I think for us, it’s not so much about losing sleep as much as [it’s about] how much can I get done in a day? It’s about, can you prepare enough to get there?”

CORRECTION: An earlier version made reference to UBS Asset Management’s Quantitative Evidence & Data Science unit, when it was actually the bank’s Evidence Lab’s experience that was drawn upon to build out the asset manager’s alternative data program.