Our approach to initial allocations

Volatility drives our allocation decisions, risk reporting, and portfolio construction. Here’s a closer look at the methodology behind our approach and how it supports smarter automated allocation defaults across every portfolio.

In this article

1. Introduction
2. The problem: one-size-fits-all defaults
3. How volatility-based allocation works in practice 
4. Why this approach

Introduction 

Ask five people what “volatility” means and you’ll get five answers: the standard deviation of last year’s returns, the VIX, the option-implied vol surface, a GARCH forecast. None of those is quite what we use. Our estimate is a factor-model forecast. We decompose each fund into a small set of well-understood building blocks, attach a forward-looking risk assumption to each one, and add them back up.

Volatility is central to what we do at Betterment. It determines the initial allocation the platform suggests when an advisor sets up a new goal, the risk category displayed when an advisor selects a portfolio for an account, and the Betterment algorithms used when constructing portfolios in the first place. When we quote a fund’s volatility, we're not reporting what happened last quarter. We're providing our best estimate of what to expect going forward, given everything we know about how that fund is built.

We’ll look at how this powers advisor workflows in practice—the problem it solves, how the mapping works, and why the approach is designed the way it is.

The problem: one-size-fits-all defaults

Betterment’s auto-glide eligible portfolios—like Betterment Core—have long used automated mapping logic that looks at a goal’s purpose and time horizon to provide a default risk level. However, that logic relied on a straightforward stock-bond split, which meant it could not extend to every portfolio type we offer.

Custom Portfolios can include up to 25 discrete risk levels, typically ordered from most conservative to most aggressive. Mapping a client to the right level requires weighing multiple signals — time horizon, goal type, and more — and that complexity grows as the number of available portfolios expands.

The same gap applied to third-party models that include invested asset classes beyond stocks and bonds. A portfolio with 70% stocks, 20% alternatives, and 10% bonds has a very different risk profile than a 70/30 stock-bond portfolio, even though both have the same equity allocation. Matching on stock-bond percentage alone would not produce a meaningful result.

We needed a single, asset-class-agnostic measure of risk that could serve as a common language across Betterment-constructed portfolios, third-party models, and advisor-customized models alike. That measure is volatility.

How volatility-based allocation works in practice

Our volatility estimate is not just an internal number.

Mapping allocations through the glidepath

When an advisor creates a new investment goal on Betterment, the platform uses the goal type and time horizon to identify a point on our allocation glidepath—the curve that determines how aggressive or conservative a portfolio should be at any given distance from the target date. That glidepath produces a Betterment Core risk level (for example, a 60% stock allocation for a retirement goal with a 15-year horizon).

For Betterment-constructed portfolios mapping directly to the glidepath is straightforward. But for custom portfolios and third-party models—which may hold alternatives, commodities, or non-standard fixed income alongside stocks and bonds—a stock-bond split is no longer a meaningful measure of risk. This is where our volatility estimate becomes essential.

Volatility as the common language of risk

Instead of matching on stock-bond percentage, we now calculate the expected volatility of every risk level in every portfolio strategy we offer, and match on volatility. The system uses account-level signals—goal purpose, time horizon, account type, and where applicable, age and income—to infer a target volatility, then selects the risk level within the chosen model whose volatility is closest to that target.

Example: Suppose a goal’s purpose and time horizon correspond to a point on the glidepath where a 60% stock Betterment Core portfolio would be appropriate. That Core allocation has a volatility of approximately 12%. The advisor has selected a third-party model whose “Moderate” risk level has a volatility of 14%—the closest available match to 12%. The platform displays the Moderate risk level as the default allocation.

Consider an account with a long time horizon and a retirement purpose. The platform recognizes the long horizon and displays the aggressive risk level, which is better aligned with the goal and time frame from day one. The advisor remains responsible for verifying the portfolio allocation is suitable for their client's goal, but the starting point is now informed by the same quantitative framework that drives our Betterment Core allocations.

Where this applies

The volatility-based mapping is now wired into key advisor workflows, including creating a client envelope and assigning a model, editing a portfolio strategy, and adding a new account. It applies to custom portfolios and eligible third-party models. For portfolios that already relied on a different, well-understood pattern, behavior is intentionally unchanged: auto-glide eligible Betterment portfolios continue to use their existing stock-allocation-based logic, and single-risk-level portfolios continue to skip the allocation screen entirely.

Advisor control and transparency

An important distinction: this is not a recommendation—it’s smarter defaulting. The advisor bears the fiduciary responsibility for the allocation, and the platform is designed to support that relationship, not replace it. When the system selects an initial risk level, the advisor sees clear disclosure explaining that the selection was based on the client's age, goal term, and other relevant factors — along with a prompt to review and confirm the allocation is suitable. They can change the risk level at any point using the same slider or dropdown they have always used.

For advisors, it means less manual work and a more scalable way to manage portfolios across their book of business. Advisors can now rely on the platform to provide a smarter first pass at risk-level assignment.

Why this approach

We designed this volatility estimation framework around four principles that guide our quantitative work at Betterment.

• Parsimony. Fourteen well-defined factors are easier to reason about, stress-test, and communicate than a 500×500 sample covariance of individual funds. The small factor set makes the model auditable and keeps the estimation problem well-conditioned.
• Stability. Forward-looking Capital Market Assumptions do not whipsaw the estimate every time markets have a rough quarter. Advisors and investors can trust that the volatility numbers they see today will not look dramatically different next month simply because realized returns shifted.
• Transparency. Every fund’s volatility can be fully attributed back to a small number of interpretable factor exposures. If an advisor wants to understand why a fund has the volatility it does, they can look at its loadings and see exactly which market risks are driving it.
• Consistency. The same CMA factor covariance matrix feeds both our volatility reporting and our algorithm. There is no disconnect between the risk we show and the risk we are managing in the portfolio.

Together, these four properties mean that when an advisor sees a volatility number on Betterment—whether on the allocation screen when setting up a new goal, in the risk summary of a portfolio, or in the details of an individual fund—that number reflects a considered, forward-looking view of risk that is grounded in the same framework we use to build and manage portfolios.

What's next

Our goal is straightforward: to help every account start in the right place on day one, backed by a volatility framework that is transparent, stable, and grounded in the same assumptions we use to build portfolios in the first place.