Saturday, December 23, 2017

Joshua Peck looks back at's evolution

Part of the credit for surprisingly strong enrollment in the ACA marketplace this season probably should go to the smooth functioning of, the federal online marketplace that serves 39 states.

In the runup to open enrollment, market watchers wondered whether hostility to the ACA at the top of HHS and CMS would extend to the website's functioning. Three weeks prior to kickoff, though, I noted that the information flow from the home page was effective, steering people both to the plan preview tool and to a quick determination whether they'd likely be eligible for Medicaid or subsidies.

On November 1 (launch day) the "see plans and prices" option disappeared from the home page. The "see plans" tool  enables a user to anonymously enter a handful of data points (zip code, ages of household members, income) and see a complete menu of available health plans, with prices that include the user's estimated subsidy. Its near invisibility led (some weeks later) to  Twitter chatter about possible sabotage. I noted that the preview tool had also been submerged in previous open enrollments -- which had baffled and frustrated me from early 2015 on. At which point Joshua Peck,'s Chief Marketing Officer before the Trump administration took office, weighed in:
 Dec 9MoreWhile counterintuitive, more consumers are more likely to ultimately find a plan and enroll if they go through the app vs. See Plans. It's not an evil plot, though when I look for it myself it can feel that way.  Direct messageThat
That led to a phone conversation. The "see plans and prices" feature is a terrific tool.  It takes less than a minute to get a sense of what you need to pay for coverage. Since multiple surveys show that large percentages of the uninsured do not know what kind or degree of financial help is available to them, the plan preview tool has always struck me as particularly valuable.

"Start here" (not there)

Peck said that he initially shared that assumption. Most visitors to can find plans for under $100 a month. "I would have expected that realizing plans can be that cheap would be game-changing."

But analysis of site visitors' behavior suggested otherwise. Peck began work at CMS two days before the ACA's second enrollment period began in November 2014. That year, the preview tool was prominently displayed throughout the enrollment period. In year 3, user testing showed "a fairly small, incremental improvement" in enrollment among site visitors who did not start with the shoparound. Those who didn't were "maybe 5% more likely to enroll. That's statistically significant -- and was replicated in a different enrollment period."

I asked whether people who use the preview tool and don't enroll may later come back and do so. That's accounted for, Peck said. "When they come, we drop a cookie, and if they come back within open enrollment, we can see whether they later start that enrollment."

Another test produced a result that was more effective than the plan preview tool, Peck said:  A green arrow on the right side of information screens prominently marked "start here." Clicks on that arrow, Peck said, "improved the likelihood of site visitors to enroll more than anything else we tested,"  (When I looked at this tool as we spoke,  it provided three options: begin an application, see plans and prices (the preview tool), or check if you're eligible for Medicaid.  At this writing, open enrollment is over, and the feature is gone, so I can't review it.)

The enrollment process itself may not be much of a barrier, Peck suggested, because it was radically simplified after the first enrollment season. By the time he arrived at the beginning of OE2,  tons of testing of user behavior and reaction had already been done. The number of screens a user had to go through to enroll had been reduced from 76 to 16 -- a talking point I recall reading at the time. Once people start an application, he said, 80-90% complete it.

The price of anonymity 

Why doesn't use of the preview tool boost the likelihood of enrollment? The "why" is somewhat speculative, Peck cautioned. But he cited several likely factors.

The first is the anonymity that users of the preview tool are able to maintain. That's important to people who may be wary of giving information to the government. But because the preview process is anonymous, "you get to the end, and there are barriers to transitioning to enrollment."

In ecommerce in general, Peck said, "it's unusual to have large quantities of people coming to look at a product and then not capture them. Anonymity is the barrier."

Because of the anonymity, those who go through the plan preview process essentially have to start over. "Window shopping is almost the same as the enrollment process," Peck noted. "The application is a much longer process -- but window shopping requires users to do two processes."

And because the preview process is anonymous, those who go through it and don't proceed to enrollment miss the crucial email followup that those who create a login but don't complete the enrollment process in one sitting receive. "That outreach is most important." Peck said. "Window shoppers don't get the reminder emails."

As someone who established a login long ago for research purposes, I can attest that the followup email chain is incessant. No one who's created a login and has not unsubscribed can possibly forget when enrollment season is on and when it terminates. (Unless, that is, they created an email address specifically for enrollment and don't habitually use email -- which is the case for many low income enrollees.) While the messaging has been more deadpan this year -- not stressing affordability, which Peck said was always effective -- it's gotten across the basics, e.g., the enrollment deadline.

HealthSherpa's single track

It should be noted that from a technological point of view, anonymity in the plan preview process should not preclude carrying the input information over into an application, once the user makes a conscious choice to start one.  HealthSherpa, a commercial online broker that sells marketplace plans only, does this. HealthSherpa is one of dozens of brokers that has permitted to establish its own interface on, so that it can process subsidies.  Back in February 2014, Michael Wasser, a co-founder of HealthSherpa, explained to me how the service avoids the "two processes" problem:
HealthSherpa claims to eliminate redundancies at every stage of the process. According to Wasser, on,  "an individual will browse to find a plan, try to sign up, and then browse to find a plan a second time -- and all the while a large quantity of redundant information is presented to the individual. We've removed the redundant step of finding a plan twice as well as optimized the information we present to the user to be as compact, specific to the individual, and easy to understand as is possible."

When customers begin at HealthSherpa, they go through a short preliminary subsidy eligibility determination. Once they begin the buying process and are jumped to (that is, the Health Sherpa interface on the federal site), the application takes into account the information they've already entered, e.g., the specific plan they' selected, and thus, Wasser said, "a lot of information they push onto every person [on the general site], that's only relevant to certain people," is eliminated on the HealthSherpa interface. "We show them the bare minimum they need to understand what's going on and feel comfortable," Wasser said. The application is then routed back to HealthSherpa, at which point the buyer sees the exact premium and subsidy and completes the transaction.
I asked Peck in a followup message whether there's a technical barrier to porting info from the preview tool into an application. He replied, "I don't think so. In OE1 [the first enrollment season], there was an angry outcry that there wasn't an 'anonymous' shopping option. My sense is that that lesson might have been over-learned."

There's always the possibility, too, that some who use the preview tool don't like what they see -- discovering that they're ineligible for subsidies, or that  subsidized plans cost too much, or that premiums are too high given the large out-of-pocket costs that those who are not eligible for strong Cost Sharing Reduction subsidies are left with. Having invested time in the application process might tip the balance for someone troubled by cost or otherwise ambivalent -- especially given the subsequent email barrage.

Time to re-test?

This year, plan preview results  might look more attractive than in the past, given the widespread availability of free bronze and cheap gold plans for subsidy-eligible buyers (a paradoxical effect of Trump's cutoff of federal reimbursement to insurers for CSR costs*). I asked Peck whether these changes might merit a review of the decision not to foreground the preview tool.

"Given the differences in how pricing works this year, there's good reason to retest," he responded. "The question would be: is the difference significant enough to outweigh the benefits of putting people through the application?"

What a long, data-intensive trip it's been

We discussed other features that had evolved over time, such as the means of steering CSR-eligible users toward silver plans (CSR is available only with silver), and the moving of searches for plans including particular doctors and medications out of the initial preview, into a filtering option once plans are available. In all, I noted that has evolved into a pretty user-friendly site. That triggered Peck's enthusiasm for the iterative practice of A/B and multivariate testing -- getting feedback from millions of site visitors, and using it.

" may be the largest portal through which people enroll in health insurance," he said. 'The site has put front and center how to change the customer experience for people who've never enrolled in health insurance before.  There's been a mini-revolution in how people apply for insurance online."

I threw a bit of cold water on this by noting that  Europeans have been buying health insurance via online exchange for some time (and so have a relatively small number of Americans, e.g., on eHealth, which pre-exists the ACA marketplace). Peck acknowledged this, but also noted that at this point, private, state-based and the federal site all have the opportunity to learn from one another, so evolution should continue.

I once saw the late healthcare economist Uwe Reinhardt, who habitually took on an air of bemused astonishment when speaking of American healthcare arrangements, walk an audience through the selection process on a Swiss healthcare exchange. Other than price, there were, as I recall, just two variables: two or three choices of deductible, and HMO vs. some kind of broader network.  In some ACA markets, in contrast, a single insurer will put up a dizzying area of minutely-differentiated plans. Other markets suffer from no competition at all. And all suffer from high prices, high out-of-pocket costs, and narrow networks.

Those market flaws are not attributable to, however. The site really has come a long way. It would be a great loss of acquired market wisdom were it to wink out, a victim of implacable Republican hatred.


*In brief: with insurers forced to price in CSR, most states instructed or allowed them to concentrate the extra cost in silver plans, since CSR is available only with silver plans. That raised premium subsidies (set to a silver benchmark) and made bronze and gold plans available at a discount. See this post for a look at how the bronze-vs.-silver choice changed in 2018 in 5 counties with the highest ACA marketplace enrollment.

1 comment:

  1. Great article. Thank you. A/B testing is better than my uninformed guessing.