Why Mint.com for Health Is a Terrible Idea, and How Keas Pivoted to the Fun Stuff

If you’re a hammer, you just want to smash nails; if you’re a programmer, you just want to build features. But features do not a successful product make. This is the central myopia that eventually blinds even the most brilliant engineer-entrepreneurs, unless they’re smart enough to surround themselves with people who can check their bias.

If you want an interesting example of this phenomenon, look no further than Adam Bosworth, the co-founder and chief technology officer at San Francisco-based health gamification startup Keas. There’s no question about this guy’s brilliance. At Citicorp in the late 1970s, he invented an analytical processing system that helped the bank predict changes in inflation and exchange rates. At Borland, he built the Quattro spreadsheet, and at Microsoft, he built the Access database. He was one of the first to propose standards for XML—the foundation of most Web services today. At Google, he helped to develop Google Docs before moving on to start Google Health.

But as everyone knows, Google Health was a failure—and so was Bosworth’s next effort, Keas, at least until the venture-backed startup went through a dramatic pivot in 2010. How Bosworth figured out that his old approach wasn’t working, and how Keas reinvented itself as a provider of health-focused games for large employers, is the tale I want to tell you today.

It’s looking like there will be a happy ending: Keas (pronounced KEY-us) is bringing on 90,000 new users per quarter and has grown to 20 employees, thanks to continued backing from Atlas Venture in Cambridge, MA, and Ignition Partners in Bellevue, WA. But to hear Bosworth tell the story, things were touch and go for a while, and Keas didn’t really turn itself around until Bosworth stopped looking at his beautiful software code and his analytics dashboards and started listening to young psychology majors and game designers.

“Most software people don’t start by thinking about psychology,” Bosworth says. “Most software people think about features first, because they are concrete and they know how to implement them. They think, ‘I would want this, therefore my users would want this.'” But sometimes—perhaps most of the time, Bosworth argues—they’re dead wrong.

Keas CTO Adam Bosworth

Bosworth grew up in New York and graduated from Saint Ann’s, a private academy where his father, Stanley Bosworth, was the founding headmaster. He says he discovered early on that he is dyslexic, and that he learned to compensate by thinking in pictures. This gave him a talent, he says, for “basically taking Lego blocks for adults, and finding really simple ways to help people build solutions to hard problems.” Those skills enabled him to make breakthrough after breakthrough in the software world, and turned him into one of the hottest commodities in Silicon Valley—Bosworth has ducked recruiting attempts by Facebook’s Mark Zuckerberg, among others.

But it was during the Google Health project that the limitations of Bosworth’s data-centric point of view began to show through. The idea behind Google Health was to get millions of people to put their health records—medications, lab results, immunizations, chronic conditions, and the like—on the Web in a central, secure repository accessible to them and their caregivers. “The idea I had was that in order to help anyone be healthier, you would need their health data,” he says. “This was in 2006, when only 10 percent of doctors had access to electronic health records, and only 10 percent of them would share it with patients, meaning that 99 percent of people weren’t able to get their own health data electronically.”

At the same time, coincidentally, personal financial management startup Mint.com was getting off the ground. “I had in mind doing Mint.com for health,” says Bosworth. Mint had three features that Bosworth wanted to emulate: reminders of bills about to come due, pretty color-coded charts illustrating the user’s progress toward financial goals, and personalized advice on ways to save money. He says he figured that if he could get users to enter some of their own health data, while grabbing the rest from pharmacy and medical lab databases, he could build a kind of health dashboard for average consumers.

Google opened up the service to great acclaim in 2008. But while it generated a lot of discussion in the healthcare industry, it never collected enough users to make it an interesting business. “It became clear to me at Google that nobody would want this,” Bosworth says. “There was nothing actionable in it. You pull the data together and you just feel confused and stupid.”

And there was another problem with Google Health: co-founder Larry Page, now the company’s CEO, didn’t like the idea. “At Google, you are free to build something Larry wants, or to leave,” Bosworth remarks. He hastens to add, however, that Page is “an extraordinarily smart guy, and in all fairness he turned out to be correct.” (Google announced this summer that it will pull the plug on Google Health on Jan. 1.)

Bosworth left Google in late 2007 and started thinking about how he could improve on Google Health. At the time, he was still enamored with the Mint.com approach to personalized data visualization. And he was absolutely convinced that Web-delivered health advice could help blunt long-brewing public health crises around conditions like obesity, diabetes, heart disease, and depression.

“Back in 1985 there was no state in this nation which had more than 10 percent obesity,” he notes. “As of 2005 there was no state that had less than 20 percent, and many that had more than 30 percent. If you want to understand why we have twice the healthcare costs of other industrialized nations, some of it is due to inefficiencies and inequities in how we deliver care, but most of it is just due to the fact that we are fatter than anyone else.”

Bosworth co-founded Keas in 2008 with George Kassabgi, the former CEO of Boston-based security software firm Bit9. With funding from Atlas and Ignition, they set out to build a system that would send consumers Mint-like text messages and e-mails containing reminders and advice tailored to their personal health goals, such as controlling high blood pressure, cholesterol, weight, stress, or diabetes. The company lined up healthcare providers such as the Joslin Diabetes Center to help design the care plans, and it built an elaborate software back end—“this huge platform with APIs and tools and all sorts of mission-critical control and deployment models”—to serve them up, Bosworth says.

Keas rolled out the technology in October 2009. But within months, a couple of big problems became apparent. First, consumers weren’t willing to pay for the care plans. To get around that problem, Keas began marketing the technology to employers—especially big, self-insured companies with a vested interest in improving the health of their workforce and thus lowering healthcare costs over the long term. Pharmaceutical giant Pfizer was one of the first to buy in.

But the second problem was more vexing: employee participation in the plans would start off high and then drop drastically. To illustrate the point, Bosworth grabbed a marker during our interview and drew a cliff-like curve on the conference room whiteboard. “It became clear we were not going to affect most people’s habits in a good way,” Bosworth says. “That brings us up to April 2010, at which point we stopped and we asked ourselves the basic question that we should have asked in the first place. Why are people unhealthy, and what could possibly motivate them to change their behavior?”

Data wasn’t the answer. The Mint-like approach, Bosworth had realized, was working more like a stick than a carrot. “All these people would enter their height and weight and lab data, and immediately we would tell them, ‘You suck. You’re overweight, your blood pressure is too high, your cholesterol is too high, you must change.’ They were gone in 60 seconds,” says Bosworth. “They know what it’s doing to their life expectancy, and they still are not doing the right thing.”

That’s when Keas finally had its “come-to-Jesus moment,” Bosworth says. And one of its saviors was Chris York, a twenty-something Stanford graduate with a bachelor’s degree in psychology and three years of training in user experience design and behavior change. “He was just a kid, an intern when he started, but he knew about behavior modification,” says Bosworth. “I said, ‘Is it possible to build something that does work?’ He said, ‘You bet.’ And I said, ‘Okay, as of now you are in charge of the user experience on Keas.'”

York was promoted to product manager. And by November 1, 2010, Keas had rolled out a completely overhauled health advisory program for 1,000 employees of its first beta-test customer, Quest Diagnostics. In York’s scheme, every Mint-like element had been removed; every piece of negative feedback was replaced with some kind of positive reinforcement. It was, in essence, a game.

“What happened was astonishing,” Bosworth says. Employee engagement rates went through the roof. Under the old system, fewer than 1 percent of employees at participating companies ever posted to Keas’s Facebook-like news feed; now 30 to 40 percent posted every week. And there was very little attenuation over time.

On the strength of those results, says Bosworth, Keas “went hastily into the process of what, in this industry, is called pivoting, which is a polite way of saying that you as an entrepreneur got it wrong, but luckily for you, you had some cash left in the bank and you can start over and get it right.”

The Keas community site developed for Quest Diagnostics

Keas’s new program works roughly like this: employees at participating companies cluster into teams of six people each, and the teams compete against each other to rack up points. Team members earn points by completing specific health-friendly actions, such as exercising five times a week, avoiding fried foods, or filling out online quizzes and health assessments. Team members can track the progress of their teammates and rival teams at their company’s private Keas portal site, where every accomplishment shows up as a post. After a set period—usually 100 days—the winning team gets a cash prize or some other incentive, and the game starts over.

At Pfizer, where 1,600 employees participated in a 12-week pilot test of the Keas program, 33 percent of participants posted to the Keas portal’s social feed—about three times the average participation rate for enterprise collaboration tools, according to the company. At the beginning of the test, only 15 percent of participants said they engaged in healthy behaviors like not smoking, exercising five times a week, and eating at least five servings of fruits and vegetables a day. By the end of the Pfizer test, that had risen to 35 percent.

Bosworth attributes such results to simple psychology. For every accomplishment—every swim, quiz, or yoga class—the game offers positive feedback. “Games are basically dopamine,” Bosworth says, referring to one of the endorphins that produce a sense of well-being. “If you have an endless series of attaboys, you have a sense of steady progress, and people like knowing that they can make progress. The game never gives you negative feedback.”

On top of that, the six-member teams are small enough that everybody’s contribution counts, which brings peer pressure into the equation. And companies don’t force employees to join—rather, they seed the contests by getting human resources employees and key influencers on board first. As Bosworth puts it, “Who are you going to listen to more, some strange company you’ve never heard of before, or one of your senior coworkers?”

While Bosworth’s original concept for a Mint-style health dashboard may still sound enticing to the geeks of the world, including adherents of the so-called Quantified Self movement, Keas has completely abandoned the idea. In fact, not a single line of the company’s original code remains in the new game-centered product. For the most part, Bosworth says, the idea of continuous self-measurement only appeals to people who are fit already, or who have an analytical bent.

“Mint for health may sound cool to you and to Esther Dyson and to every doctor,” he says. “Unfortunately it’s just fundamentally flawed in terms of basic psychology.” At the companies where Keas is engaged, Bosworth says, the average employee has a body-mass index of around 28—well into overweight territory. “If you have a population like that, it’s hard to argue for the Mint model, because there is nothing you can tell these people that is good news … most people who have Type 2 diabetes have it because they didn’t want to think about this stuff.”

Based on the success of its pilot tests in 2010 and early 2011, Keas has attracted new customers like Delta Dental, Novartis, Progress Software, and Salesforce.com, and is now rolling out its program to 90,000 employees per quarter, Bosworth says. The starting price for the program is around $30 per employee per year.

For companies, the financial incentives should be obvious: statistics show that employees with a body-mass index above 28 cost their employers an extra $2,000 per year in healthcare expenses. But even if employees don’t lose weight, they report lower absenteeism and heightened feelings of well-being after participating in a Keas contest, Bosworth says. “We don’t have data yet on how much we are going to save these companies,” he says. “But the good news is that we can be a Salesforce.com-sized business just in terms of [producing] happier, more productive employees.”

To supplement Chris York, Bosworth says, Keas recently hired another young game designer. Bosworth says he doubts the company could have pulled off its pivot without the help of these representatives of the Facebook generation. “We are betting on the kids,” he says. “George and I can try to retrospectively educate ourselves, but our sensibilities and our experiences are not competitive advantages here.”

That said, Bosworth hasn’t given up his old quantitative ways. He says he still runs plenty of queries against Keas’s customer records, checking the company’s progress and scanning for patterns. “I am very good at looking at the data, and occasionally one of these kids will assert something to be true that isn’t. I will quickly discover that, and an animated debate will ensue. My job is to trust but verify. I may know how to build Lego blocks for adults—but the point is that that has nothing to do with Keas’s survival. Luckily, I also know how to build teams of good people, and how to listen to customers.”

Wade Roush is the producer and host of the podcast Soonish and a contributing editor at Xconomy. Follow @soonishpodcast

Trending on Xconomy

By posting a comment, you agree to our terms and conditions.

  • Thanks for this great report on Keas, the most detailed and useful I’ve seen. I’ll take issue with a couple of peripheral points, since Quantified Self is mentioned in passing as a stand in for the “dashboards are wonderful magic tools that will make everybody compliant” fallacy. I completely agree this is wrong, by the way; I only take issue with the assumption that this is the approach favored by people who have been collaborating and sharing their knowledge at Quantified Self.

    We are not software producers or app makers, nor should you look at us to do interface design, though many people who come and present have pretty good skills in these areas, and have made useful, popular apps. Rather, we are advanced users and tool makers. What can you learn from us? Collaborating and sharing knowledge among pioneering users in a new field allows you to explore future, still nascent use case, and to get a bigger view of the “possibility space” than is accessible from only reading start-up pitches and business reporting.

    For instance, and relevant to the Keas story: The likely failure of dashboards has been a steady topic of discussion at QS since 2009. Last year, we posted a touching reflection about the difficulties associated with dashboards for behavior change from Marc Hedlund, whose Mint competitor, Wesabe, failed. Marc pointed out that Mint’s success in selling to Intuit, an insecure legacy software maker with a need for web services components, does not testify to Mint’s power to change behavior, and the data he has from Wesabe suggests that this change is overblown.

    The current discussion on the Quantified Self scene, which is relevant to Keas, is how easy it is to get positive test data from a well made gamification platform that has social buy in from executives and team leaders. Weight loss data from gamified, but non-technological weight loss programs with a social component, like weight watchers, is also very good. But long term data is discouraging. This is not a diss against Keas… It’s a problem for everybody to be honest about. I’m giving it here just as a taste of the higher level of discussion among the leading tool makers at QS that I wouldn’t want people to miss, because they might think we’re dashboard morons.

    Thanks Wade – !

  • Hi Gary — thanks so much for your illuminating comment. I definitely didn’t intend to reduce your whole movement to dashboards, and I don’t think Adam would endorse that view either. Quantified Self is obviously fodder for a whole separate series of articles.

    I think the overall point Adam is making, which I tried to bring out in the piece, is that data-centric platforms will probably never be very effective for behavior change in the populations that need to be reached. I asked Adam what he thought about technologies like RunKeeper, Runmeter, Withings, Zeo, Massive Health’s apps, etc. His response (which I paraphrased in the article): “I am skeptical that the people we really need to reach are the people for whom this is optimal…most of this is going to appeal people who are either very fit or very analytical in nature…I think there is a significant number of people who are very fit and very conscious, and for them this is good stuff. The problem is those aren’t the people who are sick.”

  • A couple quick points:

    1) I had always thought of Google Health as a platform without a killer app. Keas is attempting to build a killer app, but I certainly hope that they are not fully abandoning the platform dream, b/c whoever does one day succeed in on-boarding millions of users who actively contribute health data points won’t be able to write the complete universe of valuable apps by themselves.

    2) I agree with Gary’s skepticism on the long-term efficacy of “behavior change” programs (even the phrase “behavior change” is problematic). As Thomas Goetz pointed out, the gap between what we know we should do and what we do [akrasia] has vexed us for millennia.

    3) Adam Bosworth is obviously clear-eyed about these challenges, as he essentially says that he could build a giant business off making people happier – Zynga does the same thing

  • This is a great story of a company pivoting and finding something that resonates, however Adam espouses a common myth when he states “If you want to understand why we have twice the healthcare costs of other industrialized nations, some of it is due to inefficiencies and inequities in how we deliver care, but most of it is just due to the fact that we are fatter than anyone else.” Actually, a surprisingly small percentage is due to obesity (see Wa Post link below busting that myth). This isn’t to say that obesity isn’t an important issue to solve (and it sounds like Keas is making progress on that front) but there’s other bigger factors driving costs. I covered some of this in my TechCrunch posts on healthcare disruption (follow the link from my name if you are interested in those).

    Here’s the link to what is/isn’t driving why the U.S. spends so much more on healthcare
    http://www.washingtonpost.com/blogs/ezra-klein/post/why-american-health-care-costs-so-much-in-one-very-long-graphic/2011/05/09/AFjSRVbG_blog.html

  • Thanks for this comment Dave, and the link to the Washington Post infographic. I’d missed this and have definitely been hearing the statement that obseity-related diseases are driving our cost problems, I will have to go back and read your TechCrunch posts in more detail!

  • Dave — Thanks for your comment, but I am pretty skeptical about the claims you cite.

    The infographic that you mention, which was republished by the Washington Post’s WonkBlog in May, does indeed claim that “Costs in the U.S. associated with disease, including obesity, total only $25 billion in extra health care spending, a tiny fraction of the overall costs.”

    However, I think you need to look into the provenance of that infographic. It was produced by a group calling itself MedicalBillingandCoding.org, and it apparently drew on a study of medical costs that was produced by McKinsey and cited in the Incidental Economist blog (http://theincidentaleconomist.com/wordpress/the-blame-du-jour).

    If you dig into the McKinsey study you will see that the angle offered by the Incidental Economist, MedicalBillingandCoding.org, and by association the Washington Post, is a misinterpretation of the original study. That study was actually a comparison of disease prevalance and healthcare costs across several industrialized nations (Japan, Germany, France, Italy, U.K., U.S., and Spain), and it concluded that obesity and other diseases that are more common in the U.S. than in other countries account for $25 billion in *extra* costs — meaning costs relative to the costs borne by the other countries in the comparison.

    I think it’s wrong to claim on the basis of this study, as MedicalBillingandCoding.org does, that most healthcare spending can’t be blamed on disease prevalence. Which is a pretty hard claim to swallow in the first place, if you step back and consider it. If healthcare costs don’t come from treating disease, then where do they come from? Certainly not from prevention, which is notoriously underfunded by the U.S. healthcare system.

    It is probably true that obesity is not one of the most expensive diseases to treat, but it’s linked to many other extremely expensive problems, foremost among them diabetes. It almost seems to me like there’s a willful effort underway in some quarters to play down the proportions of the obesity crisis.

  • Wade – My gut sense is that obesity is the proverbial fat rat moving through the snake. I’d never argue that it’s not going to be a huge issue but I don’t think it’s the big cost driver…yet. However, my main point is that there are relatively low hanging fruit to tackle healthcare costs. The biggest thing that I have become a believer in is the Direct Primary Care model (and its counterpart Onsite Clinics) as they have shown the most impressive results at making huge impacts on downstream costs. For example, they have shown they can reduce the most expensive facets of healthcare 40-80%. These are things such as surgeries, specialist & ED visits. Led by IBM’s study of their $2B spent on health benefits, the results point to a surprisingly simple formula. More access to primary care = healthier population = less money spent.

    Both Direct Primary Care and Onsite Clinics reverse the damage that has been done to primary care models and make it economically viable and professionally desirable. See Wade – My gut sense is that obesity is the proverbial fat rat moving through the snake. I’d never argue that it’s not going to be a huge issue. However, my main point is that there are relatively low hanging fruit to tackle healthcare costs. The biggest thing that I have become a believer in is the Direct Primary Care model (and its counterpart Onsite Clinics) as they have shown the most impressive results at making huge impacts on downstream costs. For example, they have shown they can reduce the most expensive facets of healthcare 40-80%. These are things such as surgeries, specialist & ED visits. Led by IBM’s study of their $2B spent on health benefits, the results point to a surprisingly simple formula. More access to primary care = healthier population = less money spent.

    Both Direct Primary Care and Onsite Clinics reverse the damage that has been done to primary care models and make it economically viable and professionally desirable. See http://www.delicious.com/chasedave/DPCArticles for more on Direct Primary Care.