Built on Form 5500

Every retirement plan in the country, ranked by what it looks about to do

Ranked lists of the plans that look about to make a decision, and an honest count of how often each list has been right. Including where it has been wrong.

It runs on Form 5500, the return every workplace retirement plan has to file each year. Filing is not optional, so every plan is in here, not just the ones that turn up in the trade press or answer a survey.

Filing years

2009–2025

13 data files

Downloaded

3.47 GB

straight from the DOL

Lists live

2 held back

Data checks

all green

15 run on every refresh

The product, before you sign in

What a subscriber actually works from

A ranked list of named plans, each with the reason it is on the list. The grid below shows how often the top of that list turned out to be right, counted on years the model had never seen.

Plan year 2024 · 85% filed so far

3 of 100

Called in no particular order, 1 in 33 plans changes recordkeeper in the next year.

Counted over 150,904 plans, year by year, from 2019–2022, years the model never saw. The top tenth of the list did 3.19x better than picking at random.

The top of the list, and why each plan is on it

Open the full list →

1
Snow Software 401(K) Plan
Austin, TX · 10-50M · $12.0M · 58 participants
0.973
- asset mix moved sharply last year
- large share of expense going to service providers
- high admin cost per participant vs peers
2
Yuma Regional Medical Center 401(K) Plan
Yuma, AZ · 50-500M · $180.5M · 3,237 participants
0.969
- asset mix moved sharply last year
- large plan
- high admin cost per participant vs peers
3
█████████ ████ 401(K) Retirment Plan
NC · 50-500M · 848 participants
0.965
- asset mix moved sharply last year
- high admin cost per participant vs peers
5
███████ █████████ 401(K) Plan
KY · <10M · 165 participants
0.961
- asset mix moved sharply last year
- little paid to service providers
- low admin cost per participant
6
███████ 401(K) Plan
OR · <10M · 103 participants
0.960
- asset mix moved sharply last year
- low admin cost per participant
7
█████████ ████ 401(K) Profit Sharing Plan And Trust
WA · 10-50M · 7 participants
0.957
- asset mix moved sharply last year
- high admin cost per participant vs peers

Two rows are shown with the sponsor intact so the list can be checked against the DOL’s own published file; the rest are held back on this page, not hidden from you. Every row carries its own reason, which is the part a score alone does not give you.

150,904 plans scored, year by year

15,090 in the top tenth

1,450 expected to change recordkeeper

Counted on years the model never saw. Not a forecast.

Why this data is different

Nothing is missing from the list

It is not the modeling. Anyone can build a scoring model. What makes a ranking here worth something is that no plan is left out of it.

Nobody opted in, so nobody is missing

Every plan covered by ERISA files every year, or it is out of compliance. There was no survey to answer and no list to join, so there is no hole where the plans that ignored the survey would be.

Public record, not a license

The source is the Department of Labor’s own published files. Nothing here is licensed from a competitor or scraped from another vendor’s product, so nothing here can be pulled out from under you.

A plan is still the same plan after it changes its number

When a plan gets renumbered it looks like it shut down. When a provider changes its own tax ID it looks like it just won thousands of clients overnight. Sorting out who is who, year after year, is a bigger part of this build than the models are.

From the checks that run on every refresh

2,122,132 panel rows from 2,122,132/4,300,830 DC (49%) of all filers; 71% of Schedule H filers

The check confirms the filter actually removed something. A filter that quietly does nothing looks exactly like one that works, which is how the wrong set of plans shipped the first time.

What the list is worth

How often the top of each list is right

Each model learned on filings up to 2018, then ranked plans for 2019–2022, years it had never seen. The question a buyer prices a call list on is how often the plans at the top actually made the move, against a random pick and against the best sort you could build in a spreadsheet from the same filings. About 1 in 10 of the top-ranked plans changed recordkeeper, against 1 in 33 at random. The model ran 3.19x better than random; a fee sort ran 1.13x.

Signal	Top tenth of the list	Plans at random	Model vs random	Best spreadsheet sort
Recordkeeper Change Plans most likely to change recordkeeper next year.	10% · 1 in 10	3% · 1 in 33	3.19x	1.13x · a fee sort
Plan Termination Risk Plans most likely to terminate or merge away next year.	1% · 1 in 81	0.5% · 1 in 214	2.64x	1.67x · last year's events
Fee Renegotiation Plans whose administrative cost per participant is about to fall, and stay down.	33% · 1 in 3	11% · 1 in 9	2.88x	1.22x · a fee sort
First Plan Audit Small retirement plans about to cross the line that requires an independent audit.	5% · 1 in 21	0.5% · 1 in 193	9.27x	9.75x · tested_count

Built from the same data the product serves, so it cannot drift from what a subscriber sees. Each list also had to clear four tests written down before anything was measured, and every result is published, including the misses. The four tests and every result.

The one that is withheld

Built, tested, and then pulled. It is still listed here rather than quietly deleted, because if you found the flaw yourself you would have no reason to trust anything else on this site.

Pension Risk Transfer — held back: misses three of four bars

It misses three of the four bars: 0.67 on ranking against 0.70, 2.74x against 3.0, and 1.73x on the plans with no recent purchase against 2.0. It does beat the obvious rule by a wide margin. Sorting plans by how well funded they are, which every buyout pitch leans on, does worse than random here, at 0.69x. The very top is sharper than the decile figure suggests, at about 9x for the top one percent, but that is some fifty plans a year and too few to state as a result. The market itself moved under it: these transactions went from 1.3% of plans a year in the years it learned from to 2.9% in the years it was tested on, and a model trained on a quieter market ranks a busier one less well. Held back until a refresh with the newer years in training.

Provider Consolidation — held back: it was reading the answer off its own homework

It scored a perfect 1.0000 in testing, which is not a result. It is a sign something is wrong. What we asked it to predict was defined as one number crossing 0.40, and that same number was one of the things the model could see, so it simply learned where the line was and every other input scored exactly zero. Two real bugs were fixed on the way to finding it: a duplicated join that produced shares above 100%, and one provider consolidating its own registrations being read as eight separate acquisitions. The underlying series is plausible now, at 1.4 to 5.6% a year. The list still predicts nothing as it stands, and it stays held back until it is rebuilt to predict next year from what is known this year.

Who it is for

The same list, read four different ways

One firm's prospect is another firm's client about to walk. The same ranked list is a call list if you are trying to win the plan and an early warning if you already have it, so who you are decides which list leads and what a row is called.

Adviser / consultant

I want plans that are reviewing what they pay.

Win fee-sensitive plans and benchmark the ones you hold.

A row is a plan in review.

Recordkeeper / TPA

I want plans that are about to change provider.

Prospect competitors' books.

A row is a prospect.

Plan auditor (CPA firm)

I want small plans about to need their first audit.

Reach plans before their first audit engagement is decided.

A row is a audit prospect.

Incumbent, defending a book

I want to know which of my plans is about to leave.

Retain the clients you already serve.

A row is a client at risk.

Acquirer / investor

I want firms whose book is coming available.

Buy providers, not plans.

A row is a target.

Each one meets its own catch on its own home screen, not in a footnote. Read all four in full.

What this is not

Said here rather than in the small print, because these are the four things people usually assume when they land on a page like this.

Not advice. A score says a plan looks like plans that made a particular decision before. It is not investment advice, not a fiduciary opinion, and not a recommendation about any plan, sponsor or provider. A lot of people reading this are ERISA fiduciaries, so it bears saying plainly: a score on a plan is never a fiduciary recommendation about that plan.
Not a forecast. Every figure here counts what already happened, over a stretch of years the model was not built on. It is not a promise about the next stretch.
Not participant data. Form 5500 describes the plan and the employer sponsoring it. There are no records about individual people in it, and none in this product.
Not a complete year. Filings arrive nine to eleven months after a plan’s year end, so the newest year is always part filed. Every list says how much of that year has come in, rather than showing you a half filed year as if the market were shrinking.

What this is

The short version, for someone who landed here from a search result.

What does Fin360AI actually sell?: Ranked lists of U.S. retirement plans that look like they are about to make a decision: change recordkeeper, renegotiate what they pay for administration, or close. Every list comes with an honest count of how often it has been right on years the model never saw, including where it has been wrong.
What is Form 5500?: The return every workplace benefit plan covered by ERISA has to file with the U.S. Department of Labor each year. Filing is not optional, and the Department publishes what it collects as a public government record.
Why does it matter that filing is required?: Because it means nothing is missing. Every plan is in there, not just the ones that turned up in the trade press or answered a survey. A ranking only means something if it covers everybody, and most commercial lists in this market are built on a slice nobody can describe.
Is this an AI product?: No, and calling it one would claim less than the truth. The models are ordinary, well understood pattern matching over numbers pulled out of the filings. The hard part is not the modeling. It is keeping track of which plan and which provider is which from year to year, so that a plan changing its ID number does not look like a plan shutting down. The testing page explains why.
How often does it refresh?: Monthly. The Department quietly folds amendments into the same download link, so a file can go out of date without its year changing. Every refresh we ask the Department's own server what each file's size and date are now, and re-download only what has actually changed.

The rest of the questions, including the awkward ones.

The honest way to evaluate this

Start with plans you already know. Sign in and hold the ranking up against your own read on a book you can check yourself, before anyone talks about price. If it does not tell you something you did not already know, we would rather you found that out in ten minutes than in a quarter.

Data last refreshed 2026-10-18. The most recent year we can check outcomes against is 2022.