A benefits administrator at Bricklayers Local 1 Connecticut Health Fund pulled a claims report and found a $43,490 payment for a skin graft at Hartford HealthCare. The plan's negotiated rate should have been close to half that. Elevance, the plan's third-party administrator, had processed the claim at more than double the negotiated rate, and the money came from the fund's own account.

Self-fund your health plan and you control the costs. That is the line. The idea is simple: you pay claims from your own account instead of buying a fully insured policy from a carrier. You shop admin fees, you pick the TPA, and you hold the cash. If you have run a self-funded plan for any stretch of time, you have heard this pitch from brokers, from consultants, from the TPA's own slide deck. It sounds right. It is not.

The research on this is worth reading. In mid-2026, a team of RAND researchers led by Christine Buttorff published a peer-reviewed study in Health Affairs Scholar called "Paying in the Dark." They studied how TPAs price their services to self-funded employers. What they found was a structure they named "layered fee architecture."

Here is how it works. The base admin fee, the per-employee-per-month number on page one, is what employers compare when they shop for a TPA. That number can run from $5 to $60 per member per month. If it falls below $20, the RAND team found, the TPA is likely making up the gap in places you are not looking. The base fee is a loss leader. The real revenue sits behind it, built into every claim.

Sixty-seven percent of covered workers in the United States are in self-funded plans. This is not a niche problem. It is the standard model, and the fee structure runs the same way across most of it.

The fees that matter most run on percentages, not flat rates. They grow when the numbers under them grow. That is the structural flaw.

Start with out-of-network repricing. When a claim comes in from an out-of-network provider, the TPA or its repricing vendor talks the bill down. The employer then pays a "shared savings" fee, sometimes as high as 50 percent of the gap between the billed charge and the final payment. Think about that structure for a moment. The higher the original bill, the bigger the "savings," and the bigger the TPA's cut. A TPA running this model has no reason to keep billed charges low. It profits when they run high, at the employer's cost.

Then there is overpayment recovery. The TPA pays a claim wrong. Sometimes the error comes from misapplied benefit rules or contract terms, errors the TPA itself made. Later, the TPA goes back to recover the overpaid amount. Then it charges the plan a fee, typically 20 to 30 percent of the recovered sum. The plan pays to fix a problem the TPA caused. Georgetown health policy researchers gave this pattern a name: claims payment gaming. The TPA can increase its fees by letting bad payments through and then charging for the cleanup.

Then there is cross-plan offsetting. In 2017, a federal court reviewed internal documents from UnitedHealth. The court described language that "gush about how cross-plan offsetting will allow United to take money for itself out of the pockets of the sponsors of self-insured plans." The court also found that overpayments ran on fully insured plans, while the recoveries came from self-insured ones. The money moved in one direction. It moved away from the employer who funded the account.

Most people treat this as a trust problem. It is a system problem. The base fee is built to look low. The percentage-based fees are built to grow with every claim. The employer funds the bank account. The TPA controls what comes out of it.

The fix is not to stop self-funding. It is to stop treating the base admin fee as the cost. The true cost of a TPA is the sum of every revenue stream it collects from your plan, not the number on page one.

Once that is clear, three moves follow from it.

Pull the full fee schedule

Not the summary. Not the proposal deck. The actual schedule of every fee type in the administrative services agreement, including all percentage-based collections on repricing, recovery, and retained discounts. Most employers have never asked for this document in full. That is the point. It requires you to sit with a contract you signed and read the parts you skipped, the pages where the percentages live.

Run a sample-quarter claims audit

Pick 90 days of claims data. Compare the billed charge, the paid amount, the negotiated rate, and every fee the TPA collected on the spread between those numbers. You are looking for claims where the TPA earned more from the gap than the plan saved from the discount. The Bricklayers fund found one claim paid at more than double the negotiated rate. They found it because they pulled the file and read it.

Require monthly fee reporting by type

Separate the base admin fee from all percentage-based revenue the TPA collected from your account each month. If those numbers are blended into one line, you cannot see what the TPA actually costs. If the TPA resists this, that tells you something worth knowing.

Running this for one quarter does something the advice never did.

You see the gap between the admin fee you shopped and the total the TPA collected. You see which fee types sent the most revenue to the TPA. You find out whether recovery fees traced back to errors the TPA itself made. You learn whether the shared savings structure rewarded high billed charges instead of low ones.

→ What was the total dollar amount the TPA collected from your plan, across all fee types, not just the base rate?

→ Which claims generated TPA fees larger than the discount the plan received?

→ Did any recovery fees trace back to errors the TPA caused?

That is the difference between advice that sounds right and a system that proves itself.

The Bricklayers pulled their claims file and found what self-funding actually cost them. Self-funding gives you a bank account. It does not give you cost control. The gap between those two things is where your TPA makes its money.