Disclaimer

Fin360AI publishes derived analytics about employee benefit plans as institutions, for general business-research purposes. It is a prioritisation tool for people whose job is to decide which plans to look at next. It is not a system of record, not a due-diligence substitute, not a compliance product, and not a source you should cite as authority for a factual claim about any named plan, sponsor or service provider without verifying that claim against the underlying filing.

Nothing on this site or in the product is investment, legal, tax, accounting, actuarial or other professional advice, and no one here is a licensed adviser. Using Fin360AI creates no advisory, fiduciary or agency relationship between you and us, and we do not act as a fiduciary to any plan under the Employee Retirement Income Security Act of 1974 or any other law.

This section matters more here than it does on most products, so it is stated without hedging:

• No score is a recommendation. Not a recommendation to change recordkeeper, to renegotiate a fee arrangement, to terminate or merge a plan, to select or retain a service provider, or to buy, sell, hold or avoid any investment or any firm.
• No score is an opinion about a plan’s arrangements. A plan appearing high on a fee-related ranking is not a statement that its fees are unreasonable, that its fiduciaries have failed any duty, or that any action is required. It is a statement that the plan resembles plans whose administrative cost per participant subsequently fell.
• If you are an ERISA fiduciary, the duty stays with you. Any analysis you build on Fin360AI output, and anything you say to a plan sponsor or committee on the strength of it, is your own work product given under your own obligations. Presenting a signal to a fiduciary audience does not transfer any part of that obligation to us, and we have no visibility into the facts that would make a decision prudent for a particular plan.

A score is the output of a documented model applied to public plan-level filings. It is a relative statement — this plan looks more like the historical pattern than that one does — used to order a list. Specifically:

• A recordkeeper-change score is not knowledge that a plan intends to change provider, and certainly not knowledge of any discussion. It is a statement that the plan’s filing profile resembles those of plans that historically changed provider the following year.
• A fee-renegotiation score is a statement about a measurable outcome in a future filing — administrative cost per participant falling and staying down — not a judgement about whether current fees are reasonable, and not a benchmarking opinion.
• A plan-termination score currently cannot distinguish a plan that terminates from one that merges away. A flagged plan may be one someone is about to lose or one someone is about to acquire. Both warrant attention; they warrant different attention.

Every ranked row carries the reason behind its score. Read it before relying on the number: it is there precisely so that a score you disagree with can be interrogated rather than taken on faith.

Each signal is measured against four thresholds fixed before anything was measured, and it ships carrying whichever result it earned. A signal on this site may fail two of its four gates and still be published, with the failures shown next to it. That is deliberate, and it changes how you should read the output:

• A failed gate is a real limitation, not a formality. A signal that misses its top-decile lift gate is less commercially useful than one that passes it, and the page for that signal says so.
• “Not determinable” (n/d) means the result straddles its own threshold depending on the random seed. It is neither a pass nor a miss, and treating it as either is a misreading of what is published.
• A withheld signal is one that was built, backtested and pulled. It is listed rather than deleted so that the reason is on the record. A withheld signal is not a preview, a beta or a roadmap item, and nothing in the product depends on it.

Where we publish a validation figure — an out-of-time AUC, a top-decile lift, an absence lift, a base rate — that is a backtest result measured against what already happened. It describes how the model ordered a historical period it was not fitted on. It is not a forecast, not a guarantee, and not a promise that the next period will resemble the last. Model performance degrades as the environment drifts away from the one the model was fitted in; consolidation waves, rate regimes and regulatory change all do that, and at least one outcome on this site has already shown a material shift in its base rate between the fit and validation windows. Any statement about what will happen is a forward-looking statement and should be read as an opinion about probability rather than a statement of fact.

The underlying data is the U.S. Department of Labor’s published Form 5500 datasets. It may be incomplete, delayed, amended by the filer after we ingested it, or simply wrong at source — a filing error becomes our error the moment we read it. Our own processing adds interpretation: identity resolution across years, outcome definitions, and the exclusion of provider-merger events are all judgements that can be wrong. The most recent form year is always partial, because filings arrive nine to eleven months after each plan’s year end.

ALL DATA, SCORES, RANKINGS AND OUTPUTS ARE PROVIDED “AS IS” AND “AS AVAILABLE,” WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING ANY WARRANTY OF ACCURACY, COMPLETENESS, TIMELINESS, MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT.

Fin360AI is independent. It is not affiliated with, sponsored by, endorsed by or approved by the U.S. Department of Labor, the Employee Benefits Security Administration, the Internal Revenue Service, the Pension Benefit Guaranty Corporation, or any other government body. It is not affiliated with or endorsed by any plan, plan sponsor, recordkeeper, third-party administrator, adviser or other service provider named in the data. Names appear because they appear in a public filing, and are used to identify the entity that filed.

If a plan, a figure or a score looks wrong, tell us. Email admin@infinidatum.net with the plan, the field and the value you believe is correct. We check it against the panel, correct it on the next refresh if the error is ours, and tell you either way — including when the discrepancy turns out to be in the filing itself, which happens. We would much rather be corrected than be wrong quietly.

This disclaimer sits alongside the Terms of Use and the Privacy Policy. The method page documents the gates and the standing checks on the current build, and the data page documents the source, its coverage and its known limits. Where this disclaimer and the Terms of Use disagree on a point covered by both, the Terms of Use govern.