It wasn’t the outcome we wanted, but they were better off because of our work there

Vivian Graubard and Ron Gorodetzky are both former members of the U.S. Digital Service who worked together on a discovery sprint for the National Instant Criminal Background Check System (NICS) redesign in November 2015. While their story is from a few years back, the signs of disaster they saw in the project are red flags that often appear in large-scale technical implementations.

 

How did USDS get involved with the NICS redesign?

Vivian Graubard (VG): In November 2015, Mikey Dickerson, the head of USDS, went to a meeting in the West Wing, where FBI leadership informed the White House that they were launching a new NICS, which is used for background checks when someone purchases a gun. The White House was afraid that this could become the next healthcare.gov, so USDS was sent to make sure that when they rolled out the new system, they wouldn’t accidentally let a bunch of people who shouldn’t have guns suddenly have access.

Ron Gorodetzky (RG): I had started at USDS as an engineer a few months before, and what I had learned so far about government IT systems is that it’s rarely clear why they decided to build things they way they did. Usually, the way they decided to build was wrong— and often, the people using these systems every day agreed that they were built wrong.

 

What made you realize that the rollout of a new NICS could be a crisis?

VG: We went a meeting with the FBI in order to understand how the new system came to be, and they said some things that were immediate red flags. First, once they turned on the new NICS, there was no going back to the old system, even if things were going wrong. Second, none of the adjudicators who made final decisions about whether a gun could be sold had received training on the new system. Third, the contractors who built the system had not made the workflows match what adjudicators were used to, so we had reason to believe the new system would not be intuitive. Finally, they hadn’t done any end-to-end testing to see if the new NICS functioned with the other systems it interacted with, like call centers and law enforcement organizations that fed them arrest data.

RG: The system wasn’t built that way out of negligence, it’s just that OMB (Office of Management and Budget) recommended that they should buy off-the-shelf options. It’s not a bad recommendation on its face. But it meant that the gun background check system, which is by definition one of a kind, did not get custom software.

 

Once you realized how poorly designed the new system was, how did you convince people at the FBI of the magnitude of the problem?

VG: We explained that this was not a one-to-one analogy with healthcare.gov. If that website didn’t work, people don’t get healthcare, which is a failure, yes. But in this case, if gun sellers input a buyer’s request into NICS, there is a law that says the FBI only has 72 hours to give an answer. If they miss that window, then by default, the person is allowed to sell the gun. If NICS failed, it would not be felt by the public like healthcare.gov was, and the failure wouldn’t be well-understood because of how wonky the crisis was. But it had huge implications for public safety.

 

Did you find that those you spoke to in the FBI were as concerned as you were?

VG: The shooting at Emanuel African-American Methodist Episcopal Church in Charleston had happened earlier that year, and the shooter had gotten that gun because of an error in the FBI system. The adjudicators were really on edge because of that. Everybody was worried they were going to be the one who allowed the next mass shooter to get a gun.

Maybe what was most concerning, though, is that once a gun is sold to someone who shouldn’t have it, there is very little chance of getting it back. The FBI alerts the Bureau of Alcohol, Tobacco, and Firearms, and they have to do a gun retrieval. That means going to the person’s house and taking their gun, which is an incredibly dangerous situation, so ATF rarely does that.

 

Were you able to make a change in the end?

VG: Ultimately, we got them to change the date of the rollout. They had originally picked January 12, 2016, which also happened to be the State of the Union. But they had told us that anytime President Obama got on television and spoke, people went out and bought guns– probably because they had this fear he was going to take their guns away. So we painted the picture for them: Black Friday, the busiest day for gun sales, would create an enormous backlog, then gun sales would spike again around Christmas, then again during the gun town hall that President Obama was hosting on January 7, and then again at the State of the Union. So from November through January, they would be in a crunch period, and they wanted to turn on a new, untested system, losing their backlog, in the middle of that? It then became clear that they could not launch the new system that day.

RG: They didn’t launch until August, but they didn’t take any of our other suggestions. We did get money from OMB for more adjudicators, and they agreed to train them. But as far as we know, the system itself wasn’t improved. It wasn’t the outcome we wanted, but they were better off because of our work there.