Narrowing the Scope

From her position as senior vice president on the retirement plan advisory team at Sentinel Benefits and Financial Group in Boston, Julie Doran has watched big data’s trajectory. She estimates that the amount of actionable data produced by the typical defined contribution (DC) retirement plan has roughly doubled every two years for at least the last 10.

“All the signs say this data flow is set to increase even more dramatically in the next decade as the financial services industry catches up with the latest big data capabilities already in use in other sectors,” Doran says.

According to Doran, her company has moved to embrace big data in significant ways, particularly in regard to raising questions about how best to measure and promote plan success.

With the rapidly increasing visibility of real-time plan mechanics, Doran and others say, retirement industry practitioners are concluding it is not enough to just look at the overall statistics on a retirement plan—i.e., participation rate, average deferral rate, average balance and asset allocation. Using such mean data points, while informative from the employer’s perspective, can hide wide discrepancies among individuals and groups of participants, and so retirement plan advisers need to work closely with recordkeepers and potentially other third parties to utilize the data at their disposal.

Experts across the financial services industry as a whole are challenged with similar issues. Speaking at the 2018 DataDisrupt conference, Craig  Snodgrass, chief data officer of Cardlytics, said “aspirational and intent-driven data is not the same as data derived from real decisions.” Another way to put this, he said, “is that stated behavior is not always the same as demonstrated behavior, so it is important to understand how data was generated and what exactly it is telling you.”

However, when that real data is gathered, “you can take these streams of data, clean them up and extract valuable insights that look across the book of business to highlight trends and challenges that are not really visible on a case-by-case basis.”

The State of Big Data

Doran suggests it is not uncommon these days to have a client request a report from a recordkeeper or another provider—say an outsourced fiduciary investment manager or third-party administrator (TPA)—and receive hundreds of pages of data about a single retirement plan.

“That can be overwhelming, but the most important message I can share is that plan sponsors and advisers do not have to be data scientists to be able to take this data and do something actionable with it,” she says. “When you look at data specific to a given plan sponsor client, you have to get granular. Work with your providers to dig in and understand what the various groups within the plan are doing. It’s no longer enough to just look at general averages and to use this to make decisions about the design and structure of the plan.”

According to Doran, the introduction of big data technology into the defined contribution (DC) plan space “changes the ballgame entirely” for what is possible in progressive plan design, education and targeted communication—to name just a few of the areas where the expanding pool of plan data can guide plan sponsor clients’ decisions.

“Leading advisers and providers are taking the huge volume of information and distilling it into very distinct and informative points,” she says, suggesting advisers ask, and answer, targeted questions such as: “‘What are the participants who are 50 or older doing with their investments?’ and ‘What are my 20-something new enrollees doing in terms of setting an adequate deferral rate?’”

Barbara Delaney, founding principal of StoneStreet | Renaissance, a registered investment adviser (RIA) in New York City, agrees that the possibilities to better leverage data to improve plan design are practically endless. But, she points out, the refinement of plan data has also allowed plan sponsors to realize they have made some false assumptions.

“Often, with the rise of automated plans and use of target-date funds [TDFs] as the qualified default investment alternative [QDIA], the high-level averages make it look like the younger groups are doing wonderfully, while the older groups, who likely have assets outside the plan, are doing worse,” Delaney observes. “In this case and others, it is important to be able to step back and assess the quality of data and the scope of what the data is telling us. It is also exciting to see greater sharing of data to help build more accurate pictures of individual participants’ household wealth.”

Making the Most of Available Data

Delaney and Doran agree on some broad questions that all plan advisers and sponsors should ask as they grapple with new and growing troves of data: What is the source of the data being collected, and who is doing the collecting? How are they doing this collecting? Who owns the data, and how is it being polished and protected? Is this just data scraped from the 401(k) plan in isolation, or is it data that participants are interacting with and verifying in some way? Is it data that the recordkeeper is supplementing and expanding?

“While the data flows are improving so much, it’s important to always take data with a grain of salt,” Doran notes, also pointing to the appearance of older vs. younger employees’ preparedness. “This is why digging into the data, and linking different sources of data, is so important. At a minimum, you have to be parsing data by job type, age group, location, etc. This type of work adds tremendous value to the data you have on hand.”

Delaney and Doran have some sympathy for plan sponsor clients in this conversation, as the sheer volume of data can be overwhelming—and to leverage it appropriately to get the best use will require a team effort by the sponsor, the adviser and the providers.

“Using data to inform plan design is a huge element of what we are doing in collaboration with our plan clients to improve outcomes,” Doran says.
In this sense, the whole conversation about financial wellness and addressing employees’ needs beyond just the 401(k) plan balance starts with the question of how to better leverage big data.

“Another important point is that data can tell you things that you might presume are obvious[—but aren’t],” Delaney says. She cites a recent example, where a plan sponsor client wanted to make changes to its plan. “When we went back and asked how we should communicate with the 10 or so employees it has over age 70, the executives actually didn’t believe us at first that they had individuals in their work force so far beyond traditional retirement age. There are many data points you can use to help both the participants and the company.”

Doran has a similar anecdote: “I recently met with a new plan sponsor client, and I presented it with data about target-date fund usage in its plan. [Someone] said, ‘Wow, we knew our average plan age is in the 40s, but as it turns out we also have a third of our population getting ready to retire in the next five to 10 years.’ This sparked a whole additional conversation about what the plan was doing—or not—from a work force management perspective.”

Discussion of the data, in turn, led to specific changes in compensation practices to attract new, younger talent, to start to address the potential brain drain this employer faced.

“That’s just one way that looking beyond the average is critically important,” Doran adds. “Another area in particular where data can be really powerful is in helping define the appropriate target-date fund solution for a given plan population. Using big data can help you define whether a to- or through-retirement glide path is appropriate, for example.”

Cybersecurity Protection

“As fiduciaries to the plan,” Delaney says, “you need to be regularly going to your providers and asking, ‘What are you doing to protect the security of not only the assets in the plan, but also the personal identifying information and the sensitive datasets?’ You should be staying proactively aware of the ongoing enhancements of systems and processes surrounding cybersecurity. Protecting data is critical as a fiduciary.”

Not long ago, Delaney was invited to check out the so-called “cybersecurity dark room” at a very large recordkeeper, and what she saw amazed her.
“It has large teams of expert employees working all day and overnight, 24/7, whose sole job is to monitor and defend the endless number of hacking attempts and spamming attempts,” Delaney says. “Deterring the people trying to hack into our [industry’s] systems is naturally a huge priority for it. This is why we are seeing things such as voice recognition coming out, more sophisticated password requirements, etc.”

As it stands, Delaney and Doran are unaware of any major known breaches of 401(k) plan assets, but there have been some peripheral breaches of personal identifying information and some falsely acquired loans.

“We have actually heard from the recordkeepers that it is more dangerous for somebody not to have their account information set up online or not to give identifying information to the recordkeeper,” says Delaney.

“Participants might believe that staying ‘off the grid’ makes them safer, but the opposite is true. They are more of a target that may be easier to defraud, because they will be easier to impersonate, and there will just be less visibility.”

Third-Party Providers Gear Up

According to Mark Friedenthal, president and chief investment officer (CIO) of Friedenthal Financial, an RIA firm in Voorhees Township, New Jersey, the retirement plan industry is at only the beginning of the road when it comes to leveraging big data. A tool he devised to help meet the challenge is Tolerisk.

Friedenthal describes Tolerisk as “a proprietary fintech tool that assesses risk tolerance so advisers can more properly align clients’ financial and investment goals.”

“[Its] origin goes back about 10 years, when I started [Friedenthal Financial],” he says. “Like a lot of advisers, I felt like I couldn’t find the right tool or product to help clients figure out how much risk to take, in an efficient and scalable way.”

Friedenthal found himself laboring over the same kind of mathematical and data-driven techniques that he had practiced during his early career, working at some big investment companies. The work was effective but time-consuming and not at all scalable.

“In the DC plan space, in particular, we felt that the automated risk tolerance solutions were still very rudimentary and could not be relied on,” Friedenthal recalls. These consisted of using self-reported personality tests and making some basic linear adjustments to risk-taking, based on the participant’s time horizon to retirement. “We knew that wasn’t enough—it was not objective or factual enough,” he says. “To put it simply, we felt we could take the techniques we were using to evaluate bonds and bond portfolios and, instead, point these toward household cash flows.”

According to Friedenthal, this household analysis is actually much more complex than reviewing a bond portfolio. Being able to rely on advanced modeling software that simultaneously considers huge numbers of input variables and possible outcomes is essential.

A bond portfolio has pre-defined maturity durations and pays coupons at regular intervals, he explains. “On the household side, it is money going in and coming out at irregular intervals, and you are dealing with more than one individual and more than one investment horizon.

“Client response to this analysis was so powerful that, a couple of years later, another light bulb went on,” he says. He realized he could start a separate firm dedicated to analyzing data, apart from serving clients as an RIA. Packaged into an outsourced solution, Tolerisk was born.

While he is proud of the work he has done, Friedenthal freely admits his firm is one of many that are actively exploring ways to better serve DC plan advisers and sponsors with big-data-based solutions and services.

He hints at “something very big coming for the DC plan space pretty soon. Clearly, this is a very hot area of development right now,” he says.

He sees automated and verified data feeds growing only more important over time, to replace traditional client self-reporting, which, he says, “people don’t always do well. [If you ask,] ‘How much do you spend, on average, per month?’ clients will often have problems answering this accurately,” he says.
“If we can get [their] permission to aggregate their data from other tools—say QuickBooks, Quicken, Mint or eMoney—that can be so much more powerful,” he says. “This gives us a real view into the money coming in and money coming out, rather than just relying on unverified reporting from the individual.”

Art by Simone Virgini