GetCalFresh

GetCalFresh was an online helper for applying to CalFresh, California's food-benefits program. Code for America, a nonprofit, built and ran it with the state and the counties. It guided applicants through the form, let them upload phone photos of documents, and sent reminders, but decided nothing.

What it was, and what it was not

GetCalFresh was not machine learning, generative AI, or a scoring system. It was a fixed, step-by-step workflow. It had a mobile-first guided form with one question per page and an uploader for phone photos of documents. It sent text and email reminders timed to application and renewal milestones. It also had live chat in several languages and a portal for community organizations that help people apply.

GetCalFresh sent each completed application into county and state eligibility systems. There a county caseworker held an intake interview with the applicant and made every eligibility decision. The county holds this interview with every applicant before deciding. Anyone could skip GetCalFresh entirely and apply at a county office.

So its influence came through the applicant's path, never through a decision. It shaped whether people finished applications, how quickly, and whether they kept benefits at renewal.

How it grew

GetCalFresh began in 2014 as a Code for America pilot with the San Francisco County Human Services Agency, first in five counties. It replaced an online application of more than one hundred screens spread across three different county systems. By 2018 it covered more than half of California's counties.

On May 31, 2019, the California Department of Social Services adopted it as the statewide application helper in all 58 counties. The timing matched a June 1, 2019 expansion of CalFresh to people receiving Supplemental Security Income (SSI). That made roughly 500,000 older adults and people with disabilities newly eligible.

By Code for America's own account, before the statewide rollout, completing an application fell from about 45 minutes to under 10.

How big it got

Every scale figure here is Code for America's own, and none is independently audited. Its program page says GetCalFresh helped 6.2 million people obtain more than $12.8 billion in food benefits from 2017 to 2025.

Its ten-year retrospective, dated August 27, 2024, says more than seven million Californians have used the service. That counts users of the site, not people receiving benefits, so the two counts are different things.

Between 2019 and 2025, Code for America reports, more than 70% of California's online applications for SNAP went through GetCalFresh. SNAP is the federal food-benefits program, and CalFresh is California's name for it. The retrospective puts the share at about 73%.

California's participation rate rose from 66% of eligible people in 2014, fourth-lowest in the country, to 81% by 2022. Code for America presents that as context beside its growth, not as a proven effect.

What the measured effects show

Code for America and a peer-reviewed study report two effects on the applicant's path.

First, Code for America ran an in-house experiment on the SAR 7, the form people file every six months to keep their benefits. Clients got up to three reminders, and about 60% of those invited sent in the form. A last-minute deadline reminder lifted submissions among people who had not yet sent in the form from about 1.5% to roughly 12%, nearly eightfold. Code for America estimated that, used everywhere, this would mean about 315 more people renewing each month. It reports these figures without sample sizes or confidence intervals.

Second, Los Angeles County piloted flexible interviews, launched in October 2018 with full access from May 2021. Applicants could start the intake interview themselves, at a time they chose. A randomized trial designed with Tatiana Homonoff of New York University and Jason Somerville of the Federal Reserve Bank of New York tested it.

Two framings of the result exist, with different definitions. Code for America's write-up compares all applicants offered the change with those not offered it. It reports 61% against 58% approved, a difference of 3 percentage points. It reports 15 against 19 days to a decision.

The peer-reviewed version, by Giannella, Homonoff, Rino, and Somerville, appeared in 2024 in the American Economic Journal: Economic Policy, volume 16, issue 4. An earlier version was National Bureau of Economic Research Working Paper 31239, in 2023. The study covered roughly 65,000 Los Angeles applicants. Access to these interviews raised approvals by 6 percentage points, doubled early approvals, and raised long-term participation by over 2 percentage points.

The study found the intake interview is a key point where people are denied on procedure, rather than because they are ineligible. Code for America estimated the change would mean about 10,000 more households a year. The gains were largest for applicants at the county offices with the lowest approval rates before.

Why it was safe to run

In many benefits cases the danger is a decision: a fraud score, an eligibility cutoff, a risk tier, or a generated answer someone acts on. GetCalFresh made no decision at all. Every decision stayed with a county caseworker, who conducted an independent interview and checked the application against county administrative records.

The case file calls this a check set by law that sits entirely outside GetCalFresh. That is the whole safety case. Because GetCalFresh could decide nothing, it was safe to let most of a state's online applications go through one nonprofit web form.

The risk it carried

A service can be critical without holding any authority. GetCalFresh depended on philanthropy and a state partnership, not on a law. That dependency has nothing to do with accuracy, because GetCalFresh made no decision whose accuracy could be audited.

For six years no backstop was designated: no named successor ready to take over its role in helping people apply. County offices and other online applications stayed available. But if GetCalFresh had disappeared, over 70% of online applications had no planned place to go.

How it was wound down

The California Department of Social Services published a dated, phased shutdown of GetCalFresh's functions.

The SSA joint-application tool ended on September 30, 2024. SSA is the Social Security Administration. The SAR 7 renewal flow, including SAR 7 LaterDocs, ended on December 2, 2024. The application helper, the portal for community organizations, and the in-application document uploader ended on June 30, 2025. The standalone LaterDocs feature ended on September 3, 2025. The sources do not describe the joint-application tool or LaterDocs further. The data dashboards and the tools for community organizations ended on September 30, 2025.

Digital advertising, information pages, and live chat in several languages continued. The chat answered thousands of texts and emails a month statewide. Applicants are redirected to BenefitsCal.com.

GetCalFresh's design principles were built into BenefitsCal, California's permanent, state-owned portal for several benefits, rolled out from 2021 to 2023. They are plain language, mobile-first design, and simpler document uploads. So the wind-down moved GetCalFresh's role into government hands rather than simply ending it.

A contrast: Benefits Data Trust

Benefits Data Trust was a Philadelphia nonprofit that helped people get benefits. It reported more than $10 billion in benefits delivered since 2005, and more than 120,000 people served in 2023. In June 2024 it abruptly announced its closure. Its SNAP partnerships with seven states ended, and the fate of its screening tools, chatbots, and call center was left uncertain.

The Lab has a separate case on that closure. It bears on this one as a contrast: the same kind of service, removed without a successor. It is a different organization with a different funding structure, so it is not evidence about GetCalFresh.

What this network is drawn from

This network follows the pattern the case file describes. It is not a reconstruction of the actual service. It shows GetCalFresh, the assisters (staff at community organizations who help people apply) and applicants, the county caseworkers, the saved application records, and BenefitsCal. The people served, and the benefits they do or do not receive, are outside the network.

What the available tools can and cannot address

A failure pathway is a link between two parts of the network where a mistake made by one part can be passed on to the other. Here a mistake is, for example, an answer placed in the wrong field. Closing a pathway means mistakes stop passing along it. The work along it goes on.

This case has a budget of 11 units. Each tool costs the same at every target level.

Explore (No Targets) sets no targets. Under Service Targets Only, the targets are not met before any tool is used. Mistakes are copied about as fast as they are corrected. That level asks for them to be caught faster than they are copied, and for GetCalFresh to be helping the work.

Escalate checks meets that level's targets on its own, for 2 units. It raises checking when monitoring flags trouble. In all, more than 400 different sets of tools within the budget meet them.

Under Service and Safety Targets and All Governance Targets, the targets can be met, but only by spending the whole budget. Both levels ask you to close every failure pathway, among other targets. Before any tool is used, seven are open. They are Guided form to assisters and applicants, Status changes set off reminders, and Caseworker checks application and documents. They are also Answers and documents saved, County decisions recorded, Applicants' answers into the form, and Hand-off from assisters to caseworkers.

Closing all seven takes five tools. Escalate checks closes Guided form to assisters and applicants.

Mark AI-written records closes Status changes set off reminders and Caseworker checks application and documents. In the Lab it marks machine-written content in the records so readers can weigh it. GetCalFresh wrote no content of its own and used no AI. So here it stands for labeling which entries the applicant reported and which the county verified.

Gate record entries, which requires sign-off before anything enters the records, closes Answers and documents saved and County decisions recorded. Store less data closes those same two instead.

Keep prompts neutral closes Applicants' answers into the form. In the Lab it keeps questions put to an AI neutral. GetCalFresh had no AI to question, so here it stands for keeping the guided form's questions neutral.

Peer sharing rules, which sets rules for what people pass to each other, closes Hand-off from assisters to caseworkers.

So exactly two sets of tools meet the targets at these levels, and each costs all 11 units. They differ only in Gate record entries or Store less data.

Using every tool on offer, each at its strongest setting, costs 36 units, more than three times the budget. It would still meet the targets at every level.

Of the tools on offer, only Peer sharing rules starts the check named Backstop for GetCalFresh's role.