The wealth management industry has gained a new appreciation for Big Data, but digital-first firms such as Betterment have always approached clients with a quantiative, behavioral analysis approach.

The reason, according to Betterment data scientist Sam Swift, is the fitness tracker effect: today's customer wants a personalized experience backed up by figures that they can use to measure and follow their performance.

"We don’t want to hire a running coach to say, 'You looked pretty good out there today,'" Swift tells ReinventWealth. "We want to look on our GPS tracker and say, 'That's the lap I've done every month this year and I've cut it down by 10 seconds.' That's the feedback we can give people that a traditional advisor just doesn't have the tools to offer."

Swift discusses some of the technical challenges Betterment works with in providing online advice, a recently added measurement feature on its platform and the confidence gap some investors have toward robo advisors.

What doesn't the traditional industry understand about the technical aspects of what's happening in the digital wealth management space?

I don’t think there's anything that the legacy financial services industry doesn't understand about what we're doing. A lot of what we are doing — tax loss harvesting, for example — is something that's been offered for a long time. Offsetting gains with losses is a good strategy.

The thing that's new is that we’re making these types of services accessible to more people. We have no marginal costs in giving people these tools. And that's something that's new, and that's something that changes the services we can offer and to how many people; and what we need to charge people to be a viable business ourselves. So I think it’s more that we’re taking a lot of great ideas that have existed for a long time and using technology to make them available to a larger portion of the investing world.

The algorithms that are being used right now in the digital wealth management space — how do they work? And why are they so important as to how the industry is going to look in the future?

When you think about algorithms in finance, it's important to differentiate what kinds of things we're helping people to do via algorithms.

We're not an algorithm trading shop. We're not doing anything risky or speculative in terms of how we invest. Our portfolio is defined not by any sort of algorithm that's running day-to-day, but by a set of equations that decide how to properly diversify our portfolio.

What we are doing algorithmically is scaling good advice to more people. We launched a feature called RetireGuide that takes people's current situation — their prospect for saving over the next decades until their retirement, where they live, their spouses' holdings — and then gives them a recommendation today on what they should do — how much they should save and the types of investment accounts they should have.

It's not the same class of algorithmic investing that people pay a lot of attention to, but it enables us to give personalized advice, really efficiently, to a lot of people.

The traditional advisor's response is that a human can do this just as well. So why do this?

A human can do it, and it may take them a two-hour interview meeting, and four hours to go back and run the numbers, and another hour meeting to deliver the plan. And that gets the job done, but it's expensive. 

We're all living similar lives with respect to retirement. We have income now and we want to start drawing down from our savings at some point in the future, so we want to automate those basics and free up financial advisors to work on more complicated problems.

People do have estate planning and other complicated situations, but why not get the bulk of that planning work done through a smart algorithm? This is why we launched Betterment Institutional. It allows advisors to use our technology to streamline their practices.

Considering the retirement guide calculator, was that sort of advice deemed as a simple enough process to program? Is that an easy task?

I wouldn’t call it easy. There are a lot of assumptions you're factoring in, from the future of Social Security to income caps on 401(k)s and Roth IRAs to eligibility and withdrawal requirements. There are a lot of mostly regulatory limits to how you do that saving and draw down.

But it's not computationally expensive or complicated. It's the business logic around what makes a good retirement plan, and implementing that in a systematic and extensible way. We wanted to build a RetireGuide tool that works now and also works when the Roth IRA policies change, or when a new savings vehicle exists, or when we offer a new product.

What are the computational difficulties with RetireGuide and within the work that you do, as well?

I think some of the more sophisticated computations we do are around the real time features, like tax loss harvesting and Tax Impact Preview. Tax Impact Preview is one of our features where, any time you are about to conduct a transaction (a withdrawal or an allocation change, for example), we tell you what the tax implications are on that before you finish it.

That requires understanding each tax lot that you've acquired — so, for someone who has been auto depositing into a portfolio with 20 tickers over three years, we have to track the history on each of those tax lots through its entire life cycle, and then know what to sell for you if you want to withdraw something, and then calculate what the tax impacts of that sale would be.

So, it's a data management problem. Again, the actual math at the end of it is applying a marginal tax rate to the gains. But, to get there, you have to be able to track the entire history of all those financial instruments.

What's the number of Betterment users at this point?

We have more than 90,000 funded customers, with more than $2 billion under management.

Is that enough scale as a pool of users for you to work with [to do behavioral analyses]?

Sure. We want every one of our customers to have zero behavior gap to get all of the investment returns with the investment strategy that they've selected and should yield.

The analysis that I present includes tens of thousands of customers over multiple years. And, there's enough market variation in that market history, there's enough variation in the decisions they've made to learn a lot about when people make good decisions and when they're more vulnerable — problems that investment advisors have been working on forever.

We get a lot of pushback from advisors who say we'll never be able to match their ability to help people make good decisions. One of the things that I would counter from a data perspective is: How do you know how much you've helped them with your advice? That's all anecdotal and you have no ability to measure or improve on that. Whereas we can say, "We're at 22 basis points right now — average behavioral loss. We're going to squeeze that down to 15 by next quarter."

That's a form of improvement in this product that being human doesn’t help you with. You can't really measure or iterate on that success.

Is that [quantitative feedback] something you see Betterment being able to expand upon?

I think it's how many people expect and prefer to try to improve themselves and be successful — we expect that kind of quantitative feedback now. 

We don’t want to hire a running coach to say, "You looked pretty good out there today." We want to look on our GPS tracker and say, "That's the lap I've done every month this year and I've cut it down by 10 seconds." That's the feedback we can give people that a traditional advisor just doesn't have the tools to offer.

Is the confidence gap among customers with digital advice the biggest hurdle right now for fintech [companies gaining market share]?

It is one friction in terms of taking market share, but we don’t always look at it as taking market share. There are a lot of people out there who are underinvested because of the cost or familiarity burdens of getting into an advisor relationship, so they have never done it, and their money is just in a bank account. We don’t have to take market share to help all those people.

But one hurdle we face, even among people who find out about us, is that existing advisory relationships have friction. It's hard to break up with your financial advisor. You've been paying him and he's been helping you for a number of years. He's also your dad's financial advisor and it's hard to bail on that.

So, I think we've got a lot of potential out of people who get it, and are just waiting for the moment to switch into this kind of model.

Thinking about technical challenges going forward, what are the things you're looking at that you haven’t gotten to yet but you want to get to?

One thing we've been working on is that we have a portfolio of ETFs that we've chosen specifically because of their efficiency: they're highly traded, they're low cost, and they're the right way to diversify across the things that we're looking at. A lot of people have investments that are already in taxable accounts, but they can't move them to us because they're invested in other assets, and if they sell them, they have embedded gains and there'll be a tax in that transfer.

We're working toward generalizing our platform so we can accept all kinds of assets and then intelligently move people from their existing investments into our portfolio or the portfolio that we would recommend to them.

So, to lower those costs of bringing assets into our world is a technology challenge, it's not really an advice challenge.

Is it tougher to program or think out challenges for the financial industry because of regulations or these built-in legacy issues than it is if you were just doing development?

Compliance and interoperability are always part of our challenge. If you're developing a new social network, there are no rules about how it works or who wants to participate in what way.

I think that the trade-off there is that those innovations tend to be more ephemeral — so you invent a new interesting way to do a social network, everyone jumps on, but everyone gets bored and leaves two months later.

We have to make more of an investment in building something that's compatible with the rest of the financial world, but once we prove our value and people join, then we have a product and relationship that is likely to last a lot longer. So it's worth making that investment to be part of a bigger system. 

Read more: