Shoulder to Shoulder: Designing Simple Solutions that Improve Case Management System Accuracy

At Code for America, we believe the best way to improve government services is in deep partnership with the experts on the ground—the dedicated public servants who work on those services day-in, day-out. In our Shoulder to Shoulder series, we’re highlighting how these partnerships build capacity within government and improve outcomes for those who use its programs and services. Together, we’re showing it’s possible to make government work well for everyone.

The Supplemental Nutrition Assistance Program (SNAP) has long prioritized program integrity, ensuring that states and jurisdictions accurately deliver benefits to people who need them. To serve that goal, the USDA Food and Nutrition Service annually assesses each state’s payment error rate, a measure of how often a state underpaid and overpaid clients.

In 2023, the USDA imposed a $4.4 million fine on the District of Columbia due to the District’s high error rate. That’s costly for DC, and detrimental to clients. When a client is underpaid, they might struggle to put food on the table in the moment, even if they get the additional money later. When a client is overpaid, the amount they received in excess might be deducted from future payments, even if the overpayment was through no fault of their own. 

Fixing these errors was always a priority for DC, but it’s become more urgent since the passage of H.R. 1—colloquially known as the “One Big Beautiful Bill”—which introduces additional penalties for SNAP agencies tied to payment error rates. In 2030, the District could owe as much as $48 million in new penalties annually⁠—a nearly 1,000% increase. The need to use every tool at our disposal—especially modern, human-centered technology—to tackle the root causes of errors has never been greater.

Identifying where errors occur in a system

Before we design any solution, we have to understand where people are struggling within a system. We recently partnered with Washington, DC to support their effort to improve the DC Access System (DCAS), which is the case management system they use to distribute benefits like SNAP and Medicaid. 

Case management systems are inherently complex. Sometimes, it isn’t possible to overhaul a whole system. Doing so could require a huge investment, both financially and in terms of staff time, that cities and states don’t have. When we look at systemic problems through the lens of civic tech, we strive to find solutions that are effective and can make immediate impact while still working toward long-term change. 

It’s often simple solutions that end up materially improving the lives of people who interact with case management systems without requiring large, costly changes. These solutions are achievable, impactful, and can ladder up to bigger changes.

To understand where issues were occurring, we conducted more than 60 shadowing sessions where we sat with caseworkers one-on-one as they processed a case, and we kept hearing these caseworkers mention the same major pain point:  addresses. 

Addresses are critical for determining an accurate amount of benefits. To have their case processed correctly and ultimately receive the right amount, all household members on a case must live in the District, as determined by their address on file. But addresses are tricky to manage. They show up in multiple places in DCAS for multiple benefits programs, and the address on file for each program must be exactly the same. For example, if the address is correct in a client’s personal record but incorrect on the client’s income support page for SNAP, that lack of alignment could lead to an error. If a client moves and updates their address via the online portal, the system only updates the address for the primary account, but not for anyone else in the same household. In some cases, a client without access to the internet may call to update their address and later find out that a data entry error resulted in them having an incorrect address on file.

There are infinite complications. If there are 10 people on a case, and three of them have duplicate addresses, and two of them are missing an address entirely, a caseworker tasked with manually checking each person’s account for each benefits program DCAS manages faces a major challenge. We saw some caseworkers sit for an hour on a single case, tracking each person’s information on a piece of paper to ensure it was consistent across all the places it appears. The cognitive load of this task can lead to errors, which can mean delayed benefits for clients, and then later administrative penalties.

Building a simple solution to match the problem

We use a UX method called contextual inquiry when we’re trying to figure out a solution to a problem like this. We sit with caseworkers and ask them to narrate their process to explain what might otherwise be implicit to them. This exposes the underlying assumptions and processes that might not be immediately evident to a silent observer. Through all this, we realized that the cognitive load of tracking addresses is immense, and a solution that reduces that cognitive load might be an effective way to speed up case processing and reduce errors without having to make any backend changes to the system.

With this information, we proposed a low-tech, but potentially high-impact solution: a browser extension.

Our engineers got to work creating a simple prototype that we could build in less than a month. When caseworkers install the browser extension, it runs alongside their process in an application and produces a pop-up that flags potential errors. It alerts caseworkers if there are duplicate, mismatched, or missing addresses on a case and asks them to review these potential errors. It’s a simple technology that doesn’t change anything in the file, and leaves it to caseworkers to make decisions with that information.  

In our discovery process, we found that one of the most frequently occurring errors was related to “household composition,” wherein there were household members erroneously included or excluded from a household during case determination. Because of this, the browser extension also flags if people on the same case have different living arrangements than what is expected (since SNAP requires clients in the same household to buy and prepare food together at least 51% of the time) or if anyone is uninentionally omitted from the household meal group, a category that DCAS uses to determine who is included in the SNAP household.

I use it every day all day. It really is a helpful tool when you’re processing a case. Your eyes get tired after tracking so much information, and the tool lets you tackle things faster and more accurately.

We ran a four-week pilot of the browser extension with seven caseworkers and four supervisors to evaluate how well this intervention might work. When we analyzed the pilot period data, we found a reduction in processing time of around 13%. And on weekly surveys, caseworkers reported the tool helped them make more accurate determinations the first time they ran eligibility. We’re proud of the strong results from this pilot, and we’re currently working together to ensure these improvements are made permanent in the DC system.

What makes simple solutions worth it

Simple solutions like a browser extension check a lot of boxes that we look for: easily deployable in existing systems, readily understandable by the people who need to use them, and targeted to a small part of a process that can make a big difference.

Approaching a big problem like this with a mindset of inquisitiveness and innovation led us to an engineering solution that could be written in a simple programming language and tested quickly to measure its effectiveness. Embracing that spirit of innovation is what makes this partnership with the District so impactful. With small changes that start where we are, together we’re building public services that are more efficient and effective.