The Bottom Line | Published in March 2015

Data-Driven Outcomes

Using data analytics to find pieces of the participant outcomes puzzle

By Rebecca Moore | March 2015

PLANSPONSOR-March-2015-The-Bottom-Line-Story-Naftali-BederArt by Naftali-BederNearly every major retirement plan provider now offers a data analytics tool that purports to help plan sponsors determine how best to improve their plan design and participant outcomes. But how do data analytics work, and how can sponsors use such tools to perfect their plans?

As the industry has increased its focus on participant retirement readiness, it has also focused on the need to understand data beyond the basic “average” measurement. Steve Jenks, head of marketing for Empower Retirement in Denver, warns that plan sponsors should be cautious of basing their decisions on plan averages. “Averages can lie to you. The plan, on the surface, may have a great participation rate and deferral rate, but there could be pockets of employees that need help,” he says.

Plan and data analytics are a means by which sponsors can take all the information from a plan and display it in a way that will make sense, Jenks says.

Analytics can show how different workers behave differently, says Charles Berman, senior vice president of digital platforms for Fidelity Workplace Clients in Boston. For example, a 23-year-old new to the workplace will not act like someone in his 50s or nearing retirement. According to Berman, becoming aware of the distinctive behaviors can be a powerful motivator for plan sponsors to implement best practices in their plan design and participant communication.

Many recordkeepers are now able to make tools available to plan sponsors that will let them create such analytics and “visualization” scenarios, says Pete McNellis, senior vice president and co-owner of dailyVest, a supplier of software services and technology to providers such as recordkeepers and third-party administrators (TPAs), in Tucson, Arizona. According to McNellis, the data collected could include participant transaction histories, current balances, the price history of individual investment options, asset class information, and the names of different money types in the plan; it could also include indicative data such as participation status in a particular month, age group, salary range and company division. The tool can go back two, three, or even seven or eight years for this information, he says.

Typically, for plan analytics, because of the sheer volume of information, it would take too long to conduct this research in real time, so the data is collected in a batch cycle monthly. The system can show plan sponsors such analytics as plan participation and investment diversification, and can “drill down” to specific participant demographics, based on some of the criteria above. “A plan sponsor can’t identify participant needs on its own unless it analyzes them one by one,” McNellis says. “With data analytics, it can more easily pinpoint which participants need help, and take corrective action.”

Generally, when analyzing their plans, sponsors can look at plan-level or participant-level data, Berman says. For example, Fidelity’s Web-based Executive Insights tool first shows plan sponsors an overall summary of data for the plan, such as participation rate, percentage of participants using guidance or taking loans, and percentage of participants within 10% of the proper age-based equity allocation. Fidelity has rolled out a new analytics measure called OnPlan, which shows how many employees are optimizing all actions to achieve their best retirement outcomes: saving, saving enough and investing correctly. The company also offers peer benchmarking, so employers in Fidelity’s database may compare their plan information with that of companies of the same size, in the same industry or in the same geographic area.

To get participant-level data, plan sponsors may put in parameters to break down the summary information by office location or division, participant age, income, tenure or any combination of these variables.

In addition to looking at how the plan and participants are faring at any given time, McNellis says, plan sponsors may do “trending” analyses to review how plan-level statistics or participant-level actions have progressed each quarter or each year.

Because plan sponsors are increasingly measuring the retirement readiness of participants, Berman says, recordkeepers can make an estimation of how many participants are on track for a certain income replacement rate in retirement.

“Income replacement in retirement is the right measure [to base plan analytics off of], because traditional measures tend to be symptoms or causes of the problem of not being retirement ready,” Jenks says. For example, if participants are on track to replace only 50% of income in retirement, it may be because their savings rates are too low or they are improperly diversified.

Berman says the next step along the data analysis spectrum is “predictive analytics.” These systems can forecast the plan’s and participants’ outcomes if certain design changes, such as implementing auto-enrollment, are made.

Ways to Use Data Analytics

Scenario 1: Deferral Rate
  • Concern: Participants are saving below the suggested 10% to 15% of salary, including both the employee deferral and employer contribution.
  • Analytics: The plan sponsor can use a data analytics tool to determine the overall savings rate of plan participants, then “drill down” by age group. This may show that the youngest employees are under-saving. The plan sponsor can also request data to show whether participants are saving enough to take full advantage of the employer match.
  • Action: The plan sponsor may consider targeted education or automatic enrollment and automatic deferral increases. If the sponsor uses auto-enrollment, it may decide to increase the default deferral rate to that which allows participants to receive the full employer match.
Scenario 2: Diversified Funds
  • Concern:  Participants are not diversifying properly or are allocating among multiple target-date funds (TDFs).
  • Analytics: Data reveal where participants in each age group are invested, broken down by asset class. The plan sponsor can input limits to show when investments are too conservative or too aggressive. For example, for the 20-to-30-year-old age group, an equity allocation below 60% is too conservative and above 95% too aggressive.
  • Action: The plan sponsor can send targeted messages to those participants whose asset allocation is outside the proper range. If investment diversification is off for several age groups, the sponsor may consider doing a re-enrollment into the plan’s qualified default investment alternative (QDIA).
Scenario 3: Income Replacement
  • Concern: Participants in their 50s are on track to replace only 40% to 50% of income in retirement.
  • Analytics: A plan sponsor concerned with a particular age group can focus on specific data to learn whether savings rates are low for this group and/or whether these participants are poorly diversified. Depending on the tool, the plan sponsor may be able to see whether these participants have additional savings outside of the plan.
  • Action: If the participants have additional money outside the plan, the sponsor may decide to do nothing. But, if that information is lacking or analytics reveal other reasons for the low income replacement rate, the plan sponsor could implement targeted education for pre-retirees, encouraging them to increase savings or better diversify their investments. The sponsor may also consider implementing automatic deferral increases or doing a re-enrollment of all participants with a default investment into the plan’s QDIA.

Data Analytics Terms

Data visualization: Taking a large set of data and displaying it in a way that simplifies it for users via charts, graphs or lists.

Predictive analytics: Using data to predict trends and behavior patterns if one variable of the equation is changed.

Real time: Data processing in which a system receives constantly changing data as it is produced, versus receiving information from a database of historical information.

Trending: Gathering information on the pattern of change in a measurement.