Every large US retirement plan reports what it spent running itself, in a public annual filing. Almost nobody reads them. We read of them and worked out what each plan charged per person — then put the number next to the plans it should be compared against.
Nothing here is our estimate, and you do not have to take our word for any of it. Every employer below links to that employer's own filing, and every rule we applied links to the document it came from — the DOL's Form 5500 datasets, the 2024 Schedule H form and the EFAST2 filing search. Datasets read ; filings located .
Median administrative expense per participant, by how many people are in the plan. Each bar is the median of every qualifying plan in that band for form year . This is the whole reason plan size matters.
Computed by us, from the DOL's
F_5500_2024_Latest and F_SCH_H_2024_Latest
files, read — every qualifying plan in the dataset, not a sample.
Method and scope are in the section below.
The largest plans we have read so far, by number of participants with an account balance. More are added each week. Every one of them carries a link to the filing the numbers were read from, so you can check the figure against the document.
The figure is real and it is the sponsor's own. It is also narrower than it sounds, and saying so is the point of this section.
Schedule H
total administrative expenses — DOL dataset
field TOT_ADMIN_EXPENSES_AMT, line 2i(12) on the 2024 form — divided by the count of participants with account
balances from the Form 5500 itself. Both are filed by the plan sponsor under penalty of
perjury and are public under
29 U.S.C. § 1026(a).
The line number is worth stating carefully, because it moved. Schedule H's expense section was restructured after the 2019 form, where the total sat on line 2i(5) over four sub-items. The 2024 form breaks the same total into twelve, and the total is line 2i(12) — checked against the 2024 form itself, read . The dataset field name was verified separately against the layout file shipped inside the dataset; the Form 5500 Datasets Guide and the 2024 data dictionary document both.
We use the DOL's "Latest" datasets, which carry only the most recent accepted filing per plan, so amended and duplicate filings do not double-count. Each plan's filing was then located by EIN and plan number in EFAST2 on ; exactly one form year filing existed for every plan on this site, and every plan name EFAST2 returned matched the name in our data.
Line 2i does not include fees netted inside a fund's expense ratio, which for many plans is the largest cost of all — it is deducted from returns before the plan ever sees it. It also excludes anything the employer pays directly out of its own pocket rather than out of plan assets.
So a low number can mean a cheap plan, or a generous employer, or costs that simply sit somewhere this line does not reach.
We do not tell you what your plan "costs you", because this figure is not that. We report what the sponsor reported, label it as such, and show the comparison. Any number we could not read from a filing renders as not reported — never as a zero, never as an estimate.
We rank nothing, recommend nothing and sell nothing.
And we do not ask to be believed. Every employer card links to the exact filing its numbers came from, by its EFAST2 acknowledgement ID. If our number and the document disagree, the document is right — please check us.