In my work as a behavioral economist, I’ve thought a lot about how nudges can drive lasting behavior change. In the domain of retirement savings, Nobel laureate Richard Thaler and I devised a program called Save More Tomorrow back in the mid-1990s that used nudges to help people make better decisions about their long-term financial future. That program invites employees to gradually increase their savings rate over time, and it has been a success: according to my latest estimates, it has boosted the savings rates of as many as 15 million Americans.

Unfortunately, it took us 20 years to help that many people. The slow pace of this process has led me to become increasingly interested in digital nudging, which seeks to identify online designs that help people make smarter choices. The advantages of digital nudging are two-fold. First, the digital space allows us to conduct research much faster, as we test out multiple designs to see which one works best. Instead of waiting years to see if an intervention is effective, we can often get results in days or weeks. Second, the digital world offers unprecedented scale: by fixing a single website or app, we can potentially help millions of people make better financial decisions.

In recent years, my colleagues and I have done more research to explore the enormous potential and cost-effectiveness of digital nudging. From the use of Big Data to improve retirement outcomes to small tweaks to the screens of a leading robo-saving app, we’ve shown that improving the design of the online world can have a big impact on our financial well-being.

Here, I’ll outline new research and case studies that help us understand digital nudging even more deeply:

Case One: The Power of Framing an Email

In a recently published paper I authored along with John Beshears, Katherine Milkman, Cass Sunstein, Richard Thaler, Maya Shankar, Will Tucker-Ray, William Congdon and Steven Galing, we looked at the impact of an email intervention on enrollment in a savings program sent to roughly 800,000 military service members. (A control group received no email.)  We did the study in conjunction with the White House Social and Behavioral Science Team, also known as the Nudge Unit within the Obama administration. The most effective email nearly doubled the program enrollment rate compared to the control group. This email used simple nudges, such as providing actionable steps for enrolling in the savings program and giving people clear examples of how small contributions can lead to large account balances, to make the responsible choice the easy choice.

Because the email intervention was relatively low-cost — the Department of Defense only had to spend around $5,000 — the nudge was roughly 100 times more cost-effective than traditional policy interventions, such as tax incentives and financial education programs. If a basic digital tool like email is so effective, I’m optimistic we can get even better results using text messaging, personalized videos, and chat bots.

Case Two: The Power of Mental Accounting on Screens

Workers used to rely on their employer for retirement, whether it was a defined benefit pension or matching contributions to their 401(k) accounts. But according to the General Accounting Office, more than 40% of American workers are no longer salaried employees: they are self-employed, independent contractors, freelancers, part-time workers and others often referred to as contingent workers. It doesn’t matter if they are driving for Uber or running a coffee shop: these people will be solely responsible for their own retirement and financial well-being. If we don’t provide them with easy digital tools for savings, we could be looking at a generation of workers struggling to achieve even a semblance of financial security in retirement.

My UCLA colleague Hal Hershfield, Steve Shu from City University of London, and I have been working with Acorns, a robo-saving app, to help their users save more money. Thanks to innovations like Apple Pay and Amazon’s one-click buying, it is easier than ever to spend money online. The goal of robo-saving apps is to make saving money just as easy.

One of the main issues we’ve tried to address is the tendency of people to not enroll in the automatic deposit program, in which a set amount of money is moved every month into a savings account. Such automated deposits are by far the most effective way to build up a nest egg: if people have to actively think about saving, then they probably won’t do it.

To see if we could help more users enroll in the automatic deposit program, Hal, Steve, and I experimented with several different versions of the same question. On the enrollment screen, users were randomly assigned to one of three categories.  Some were asked if they would like to a save $5 every day, some were asked if they wanted to save $35 a week, and some were asked if they wanted to save $150 a month. While only 7% opted to save $150 a month, nearly 30% decided to save $5 a day. That’s a huge shift in choices, especially since the amounts are all essentially equivalent. Saving $5 a day makes us think about skipping a Starbucks latte (that seems doable), while $150 a month makes us think about car payments, which is a more daunting amount to give up.

In the 21st century, the tradition of the monthly or bi-weekly paycheck is increasingly obsolete. Many workers have a highly variable stream of income.  As a result, people need a digital solution — such as the robo-saving app described above — that directly and promptly links their savings to their income, even when that income appears inconsistently. We need to make it as easy for these workers to save as it is to spend.

Case Three: The Power of Just-in-time Mobile Feedback

Consumers make more informed choices when they have the right information at the right time. Digital and mobile technology makes this process feasible on a mass scale. Along with Yaron Levi, now a behavioral economist at USC, we set out to test the impact of the Personal Capital mobile app on its users. (Personal Capital offers account aggregation service that brings together all of your financial accounts.) The star attraction of the website and app is the dashboard, which lets people track their investment performance and spending behavior. For instance, the dashboard breaks down, by category, where the money goes; it shows you how much you spend at clothing stores and the percentage of income that’s devoted to paying off the mortgage. The app allowed users to get this information on their mobile devices. While the exact same information was previously available to users on their desktops and laptops, the app made it easily accessible to people when they were making purchase decisions.

Did this mobile access change the way they spent their money? The app turned out to have a huge impact, with the average user decreasing their monthly spending by 15.7%. Not surprisingly, almost all of that decrease came in the form of discretionary spending, with users spending far less on items such as dining out. The results fit with recent government surveys, which find that the majority of consumers with access to their financial information on mobile phones check their balances before making large purchases. Of those who check, 50% decide not to buy an item because of the feedback.

Other studies show that traditional financial education is typically ineffective, often because people struggle to apply the lessons they learn in the classroom to real life months or years later. However, the just-in-time financial information provided by a good app can solve this timing issue, providing us with crucial information when we need it the most.

The Future of Digital Nudging

The above studies involved experimentation with different digital nudges in the field.  However, the future of digital nudging should also include a broader toolset, including Big Data and personalization.

One project I’m very excited about is being led by the Voya Behavioral Finance Institute for Innovation, where I am Senior Academic Advisor. The project involves using Big Data to assess whether a person is making financial decisions using their fast instincts or carefully reflecting on the alternatives. We do this by measuring basic aspects of their visit to their retirement website, such as how long they spent on the site and whether or not they explored the various tradeoffs in their decisions. We are seeing a strong correlation between online decision-making style and projected retirement income, with people who engage in more reflection also saving more money. Furthermore, once we know how a person makes their decisions, we can tailor our nudges accordingly. For example, a person who spent less than 10 seconds on the retirement enrollment website, and is projected to have a large shortfall, might benefit from greater digital nudging. Fortunately, smart software can identify those who don’t think much about their financial future and suggest effective nudges.

In addition, I’m also working on the development of short assessment tools that can measure a person’s tendencies across a variety of behavioral traits, from loss aversion to present bias. This allows us to create highly personalized financial recommendations. For instance, someone who is extremely loss averse should probably be offered a different mix of investment funds, or insurance coverage, than someone with a higher tolerance for losses.  When you visit Amazon.com, the website offers you a curated list of products based on your preferences and needs. I believe our financial lives could benefit from a similar level of personalization.

These digital nudges don’t need to only apply to the financial domain — similar principles can apply to decisions involving health, sustainability and education. What’s the best way to nudge people to exercise?  What kind of digital feedback encourages people to consume less energy?  How can we personalize online education?

The design of the digital world shapes our behavior in countless ways.  The more we can do to understand the most positive uses of behavioral science combined with digital, the better off the world will be.