Unemployment Insurance Claims Software: Why the System Only Fails When Every Worker in the State Needs It at Once

The system works fine until the day it matters

A large employer announces a plant closure on a Thursday. Eleven hundred people file over the weekend. By Monday the agency's call center is at capacity, the online filing system is timing out at the employment history step, and the adjudication queue that normally runs four days is at nineteen. Every one of those claimants has rent due, and the agency's performance is measured on first payment promptness, which is now unreachable for this cohort no matter what anyone does.

Unemployment insurance is the only public program designed to absorb its own demand shock, and the systems running it were mostly built when the shock was assumed to be a recession, not a week. The 2020 pandemic exposed this at national scale: legacy mainframe systems that had operated adequately for two decades could not stand up new federal programs in the timeframe Congress had set, and the failures were on the news for months. Most states responded by starting modernisation programs with Geographic Solutions, Sagitec or Fast Enterprises, and many of those programs are still running.

If you are inside one of those agencies, the useful question is not whether to modernise. It is which parts can be improved now, on their own, without waiting for the program.

Problem 1: monetary determination is the easy half

Establishing monetary eligibility is arithmetic against quarterly wage records: identify the base period, sum wages, apply the state's formula, produce a weekly benefit amount and a maximum benefit amount. Most legacy systems do this correctly, including the alternate base period cases that require the most recent quarter's wages before they have been reported.

The queue is on the other side. Separation fact finding requires the employer's account of why employment ended, the claimant's account, and an adjudicator deciding whether the separation disqualifies under state law. The employer response arrives late or never. The claimant statement is a free text box that produces three lines. The adjudicator then telephones both parties, which is why a determination takes weeks.

What a custom adjudication workspace does: replace the free text box with structured, branching fact finding that asks only the questions relevant to the separation type, so a voluntary quit case collects the specific facts the statute requires rather than a narrative. Employer responses arrive through the national separation information exchange with a deadline clock the system enforces and a documented consequence when it lapses. The adjudicator then sees both accounts side by side with the conflicts highlighted, drafts a determination from a decision framework that cites the statute, and moves on. Agencies that fix fact finding structure rather than adding adjudicators see the queue move, because the bottleneck was never staffing alone, it was that half the file was missing on first touch.

Problem 2: the front door became a fraud target and stayed one

Organized identity fraud against unemployment programs during 2020 and 2021 was extensive and widely documented, and it permanently changed what a claims front end has to do. Every state now runs some form of identity proofing, and the ones that bolted it on as a gate discovered the second problem: legitimate claimants who cannot pass remote proofing, disproportionately older workers, people without credit history and people whose identity was already stolen, now cannot file at all.

What a build worth funding does: treat identity as a risk score rather than a gate. Low risk claims with a clean data match proceed. Higher risk claims proceed but with payment held pending a defined verification step, which can be in person at a workforce office. High risk patterns, meaning many claims from one device, one bank account across unrelated claimants, or filings against employers with no separation reported, route to an investigations queue before any money moves. Cross matching against the national new hire directory and incarceration data catches a different class of improper payment. The design principle is that stopping a fraudulent payment before it leaves is worth more than recovering it afterward, since recovery on identity fraud is close to nil.

Problem 3: weekly certification is the highest volume transaction you run

Every claimant certifies weekly, answering the same questions about availability, work search and any earnings. It is the most repeated transaction in the agency and it is where usability becomes policy. A confusing earnings question produces an overpayment three months later that the agency then has to establish, notice, appeal and collect, which costs more than the payment.

What a custom certification channel does: work properly on a phone, because a large share of claimants have no other device. Ask the earnings question in a way that distinguishes when work was performed from when it was paid, since that single confusion generates a meaningful share of non fraud overpayments. Accept work search entries as structured records with employer detail captured at the moment of the search rather than reconstructed weeks later during an audit. Save partial progress, because certification interrupted by a shift is certification abandoned. And return a clear status afterward, since the second largest driver of call volume in every agency we have looked at is claimants calling to ask whether their week processed.

Problem 4: the tax side is a different business sharing the same books

While benefits go out, the other half of the agency runs an employer tax operation: registration, quarterly wage and contribution reporting, experience rating, benefit charging, delinquency and audit. The same employer account is charged for benefits paid to its former workers, which is why employers protest separations, and the two sides of the house frequently run on systems that do not agree about who the employer is.

What a custom employer portal does: give one account view covering registration details, quarterly filings and payments, benefit charges with the claims that caused them, and open separation requests with their deadlines. Third party administrators, who file on behalf of many employers, get proper delegated access instead of sharing credentials, which is both a security fix and a data quality fix. Charge protests are filed and tracked in the same place they are visible. This is the component that generates the least political attention and the most goodwill, because employers experience the agency almost entirely through it.

Problem 5: emergency programs arrive with weeks of notice

The federal pandemic programs were legislated and expected to be paying within weeks. Whatever the next emergency program is, the notice period will be similar, and the agencies that coped were the ones that could add a new benefit type, new eligibility rules and a new payment stream without rebuilding.

What that requires architecturally: benefit programs as configuration rather than as code paths, eligibility rules that are versioned and effective dated, and a payment layer that does not care which program generated the entitlement. If your current system requires a release to add a program, you already know how the next surge goes. This is the strongest technical argument for modular work even inside a larger modernisation, because a module that owns program definition can be built in months and used for decades.

What this costs and how long it takes

Across Digital Heroes delivery experience, a module in this domain runs $150,000 to $500,000 over 5 to 10 months. That covers a claimant intake and certification channel, an adjudication workspace, an employer portal, identity and fraud triage, or an appeals system with scheduling, hearings and decisions.

What drives the number:

• Access to the benefit system of record. If the only interface is a nightly batch, the module has to tolerate stale state and be honest with users about it.
• Federal tax information, since wage and offset data brings IRS Publication 1075 controls into the environment and that is a security program with its own review.
• Identity proofing vendor integration, plus the alternate path for people who cannot pass it, which is a policy decision as much as an engineering one.
• Languages and accessibility, which for a program serving the entire workforce is broader than most government portals and needs building in from the first screen.
• Appeals, if in scope, because hearing scheduling with parties, representatives, interpreters and recorded proceedings is its own product.

Build versus buy

Do not replace the benefit calculation and payment engine as a custom project while claims are running through it. That is what the large modernisation vendors are for, and the reason they take years is mostly not incompetence, it is conversion and parallel running against a live payment stream.

Do build the channels and the workspaces. Claimant intake and certification, employer portal, adjudication, fraud triage and appeals are all separable, all improvable within a year, and all measurable against numbers your agency already reports to the federal government. The best argument for doing this alongside a modernisation rather than after it is that these components teach you what your real requirements are before the big program freezes them.

If you are a small state or a territory with a modest claim volume and no active modernisation, buying a hosted benefits system from one of the established vendors is a defensible answer, and we would say so rather than pitch a build.

How to choose a developer for unemployment insurance work

Ask how they would design fact finding for a discharge for misconduct. If they propose a text area, they have not read a state statute and your adjudicators will still be making phone calls.

Ask what happens to a claimant who fails remote identity proofing. The answer must include a path that ends in payment, not a dead end, and the fact that they have thought about it tells you whether they understand the population.

Ask how a new federal benefit program would be added. If the answer is a code release and a deployment, you are buying the same problem your current system has. Program definition belongs in versioned configuration.

Ask who owns the code, the rules and the environments, and get it in writing before kickoff. Unemployment systems outlive administrations and vendors, and an agency that cannot move its own logic to another firm has bought a decade of change orders. At Digital Heroes the agency owns the repository from the first commit.