Unemployment Insurance Claims Software: Build vs Buy for a State Workforce Agency

What Geographic Solutions, Sagitec and Fast Enterprises actually do well

You are probably not choosing between a blank page and a product. You are choosing between joining the queue for your state's next modernisation contract and commissioning something narrower that could be live before that contract is signed. So start by being fair to the products, because the case for building anything rests on what they already cover.

Those three vendors have each carried multiple state workforce agencies through a full unemployment insurance (UI) benefits and tax replacement, and what they do well is the arithmetic and the plumbing. Monetary determination against quarterly wage records, including the alternate base period cases where the most recent quarter has not been reported yet. Weekly benefit amount and maximum benefit amount under your state formula. Payment with Treasury offset and overpayment recovery. Connections to the State Information Data Exchange System (SIDES) for employer separation responses, to the Interstate Connection Network (ICON) for combined wage claims, and to the National Directory of New Hires for cross matching. The federal reporting pack, meaning ETA 227, ETA 581, ETA 9050 and ETA 9052, comes out of the box because every one of their customers files it.

That is a great deal of unglamorous correctness and it took each of them years. If you are weighing custom versus off the shelf for the whole system, buy. We will say that before anything else.

Where the packaged systems stop: separation fact finding

Monetary eligibility is the easy half. The queue sits on the other side of it, in the nonmonetary determination, and that is the workflow the packages model badly.

A separation issue needs three things: the employer's account of why the job ended, the claimant's account, and an adjudicator deciding whether the separation disqualifies under your state statute. In every packaged system we have looked at, the claimant's account is a free text box, and it produces three lines. A voluntary quit case usually turns on whether the claimant raised the condition with the employer and allowed a chance to correct it before leaving. A discharge case turns on wilful disregard of the employer's interest and whether a prior warning exists. None of that is in three lines, so the adjudicator telephones both parties twice and the determination takes weeks.

The employer side has the same shape. SIDES gives you a structured exchange with a response deadline. What the package does with a late or missing response is record it. What your statute usually says is that a response outside the window carries a defined consequence, and the agencies clearing their queues are the ones where software enforces that consequence rather than leaving it to an adjudicator's judgement on a Thursday afternoon.

Branching fact finding that asks only the questions your separation type requires is not a feature the vendors refuse to build. It is a feature that has to be built against your statute, and your statute is not their product.

The arithmetic: cost per claim against the cost to build

Do this with your own numbers rather than ours. Hosted UI systems here are priced either as a fixed annual hosting and maintenance fee or as a fee scaling with covered workers or initial claims. Take your current figure, divide by initial claims processed last year, and you have cost per claim. Divide it again by named staff who log in and you have cost per seat. Agencies almost never compute either, and both change the conversation.

Now the other side. One module sits in the band described below. Amortise the midpoint over five years with support included and you land near $65,000 a year. That is not a replacement for your platform contract, so the honest comparison is against the staff hours it removes rather than against the licence.

The crossover in this category is a transaction count, not a seat count. A custom adjudication workspace starts paying for itself somewhere above roughly 40,000 nonmonetary determinations a year, because that is the volume at which taking four days off average time lapse frees more adjudicator hours than the build costs. Below about 15,000 determinations a year, meaning a small state or a territory, the arithmetic does not work and you should configure what you bought. Between those two numbers it turns entirely on whether your time lapse against the federal core measures is already acceptable.

What a custom build actually costs, including migration and year two

In Digital Heroes delivery experience, a module an agency can commission and actually finish runs $150,000 to $500,000 over five to ten months. That covers one of: a claimant intake and weekly certification channel, an adjudication workspace, an employer portal for separation responses and quarterly contribution reporting, an identity and fraud triage layer, or an appeals system with scheduling and decisions. A full benefits and tax modernisation is not in that band, and any firm quoting one inside eighteen months is the reason this sector has the reputation it has.

Then the two lines that never appear in a request for proposal. Data migration runs 10 to 25 percent of build cost and lands at the top of that range whenever claim history is in scope, because a determination is evidence and has to stay producible for years. Year two and every year after runs 15 to 20 percent of build cost annually. That is not padding, it is legislative change: your statute moves, your programmes move, and an unfunded system is one your staff work around within three years.

Three cost drivers are specific to this domain. Wage and offset data pulls IRS Publication 1075 controls into the environment, which is a security programme with its own review cycle. Section 508 accessibility and language access for a programme serving the whole workforce is wider than a normal government portal and has to be designed in from the first screen. And access to the system of record: if the only interface is a nightly batch, the module must tolerate stale state and say so to the user.

The four situations where building beats buying

Regulatory fit. Disqualification law, good cause standards, work search requirements and appeal timelines are all state statute. When a policy change takes your vendor two quarters because you sit behind thirty other states in their release queue, that is structural and configuration does not solve it.

Scale economics. Above roughly 40,000 nonmonetary determinations a year the arithmetic turns, and it turns harder when your first payment time lapse already sits outside the federal acceptable level of performance, because the alternative is hiring adjudicators you keep paying after the surge ends.

A workflow that is how you are judged. A public agency has no competitors but it does have a scoreboard, and yours is first payment promptness, nonmonetary determination time lapse and appeals timeliness reported to the US Department of Labor. If the queue dragging those numbers is modelled by your product as a generic work item, that queue is worth owning.

Integration sprawl across three or more systems. Count them honestly: benefits system of record, quarterly wage records, SIDES, ICON, the National Directory of New Hires, your identity proofing vendor, the debit card or direct deposit provider, the hearings platform. Once one workflow crosses three of those, the product becomes a participant rather than a hub and the coordination logic has nowhere to live except in a person's head.

One further trigger. Federal emergency programmes arrive with weeks of notice. If adding a benefit type to your current system requires a code release, you already know how the next one goes, and a module holding programme definition as versioned, effective dated configuration will be used for decades.

How to decide in a week, and what discovery leaves you owning

Run this instead of commissioning a study. Pull fifty nonmonetary determinations closed last quarter, at random rather than chosen by an analyst. For each, record five timestamps: claim filed, issue detected, fact finding sent, last party response received, determination issued. Ninety minutes of clerical work gets you the distribution.

Then read where the days sit. Mostly between detection and fact finding being sent means a routing problem, and configuration may fix it. Mostly between fact finding sent and the last response received means a fact finding design problem, and a build is justified. Mostly after the last response means an adjudicator capacity problem, and software is not your answer this year. That distribution settles the question more reliably than any demonstration.

Then pay for a discovery phase rather than accepting a free proposal. At Digital Heroes that produces a signed product requirements document before any code exists: data model, integration inventory, permissions, the IRS Publication 1075 boundary, acceptance criteria and a fixed price. You own that document whether we build it, your incumbent builds it, or nobody does. It is also what makes four quotes comparable, which is the thing agencies most often lack.

Who we are wrong for: if you want one vendor accountable end to end for benefits and tax including the payment engine, that is a prime contract and you should run a procurement rather than call us. We work as India LLP, US LLC and UK LTD entities so intellectual property assigns under your own law, the agency owns the repository from the first commit, and you meet the named engineers before signing. More than fifty specialists, over 2,000 projects delivered, checkable on Clutch, Trustpilot, Fiverr Vetted Pro and D-U-N-S.

Book a 30-minute call with Digital Heroes and get a written plan and a fixed quote within 48 hours.