Data sourced from Dakota Marketplace, the global LP and GP intelligence platform trusted by thousands of investment professionals. Learn More | Book a Demo
13F filings are public, free, and updated four times a year. That combination makes them one of the most widely referenced datasets in institutional finance.
But the question investment professionals eventually ask (usually after building a prospecting strategy on top of the data) is whether they can actually trust what a 13F says.
The honest answer is: it depends on what you mean by accurate, and what you are trying to do with the data.
13F filings are legally required disclosures, which means the positions they report are real. But between the 45-day lag, amended filings, confidential treatment requests, CUSIP errors, and the sheer complexity of processing thousands of filings correctly, the gap between a raw 13F and an accurate, actionable dataset is wider than most teams expect.
In this article, we cover what 13F data is genuinely accurate at, where the accuracy breaks down, and how Dakota's holding data is built to close those gaps.
What 13F Data Is Accurate At
Start with what the filing does well. A 13F is a legal disclosure. Every institution managing $100 million or more in Section 13(f) securities is required by law to report every long position in those securities within 45 days of each quarter end. The SEC enforces this. Filers face real consequences for material misreporting.
That means at the most fundamental level, did this institution hold this security at this quantity as of this date, the answer in a correctly filed 13F is yes. The position existed. The share count is real. The market value was calculated from a real price on a real date.
For fund managers using 13F data to identify RIAs adding alternatives or build a switch prospecting strategy, that foundation is genuinely useful. The position-level data, at its source, reflects something that actually happened.
The accuracy question gets complicated when you move from the raw filing to the processed dataset your team is actually working from.
Where 13F Accuracy Breaks Down
The 45-Day Lag
The most widely understood accuracy issue is also the most unavoidable. By the time a 13F is publicly available, the positions it reflects are already up to six weeks old. A filing released in mid-May shows holdings as of March 31. A lot can change in six weeks — positions can be added to, reduced, or exited entirely.
This does not make the data inaccurate in a technical sense. It makes it historical. The filing accurately reflects what the institution owned on a specific date in the past. Using it as a proxy for what they own today is a meaningful leap, and the further you get from the filing date, the wider that gap becomes. The implications of this for outreach timing are significant, something we cover in detail in how to use 13F data to time your outreach.
Amended Filings
The original 13F is not always the final word. Filers can, and regularly do, submit amended filings (13F/A) to correct errors or omissions after the original deadline. An amendment supersedes the original, which means any dataset that processed the original and did not pick up the amendment is working from incorrect data.
The SEC does not send alerts when amendments are filed. If a data provider is not continuously monitoring EDGAR for amendments, their users may be looking at positions that the filer itself has already corrected. The scale of this problem is larger than most people realize… amendments are common, and missing them introduces real inaccuracies into downstream analysis.
CUSIP Errors and Security Misidentification
13F filers are required to identify securities using CUSIP numbers, but in practice the raw data is riddled with errors. CUSIPs are mistyped, outdated, or simply wrong. The same ETF may appear under multiple CUSIPs across different filers. Share classes are frequently confused with each other. When a fund changes its structure, ticker, or CUSIP, filers do not always update consistently… creating phantom positions and broken time series in any analysis built on the raw data.
You cannot simply trust the CUSIP as filed. Every security identifier has to be validated against a live reference database and corrected where necessary. This is a significant processing step that separates a raw EDGAR feed from a clean, reliable dataset.
Dakota's holding data processes every 13F, including amendments, through a seven-step enrichment workflow, so your team never works from stale or uncorrected data. Book a demo.
Confidential Treatment Requests
In limited cases, filers can request that specific positions be omitted from the public filing. These requests are granted at the SEC's discretion and are relatively rare… but when they are granted, those positions are simply absent from the public record with no indication that data is missing. A filer's total reported market value may be materially understated if significant positions are under confidential treatment. There is no way to detect this from the public filing alone.
The No-Context Problem
A raw 13F gives you a firm name, a CUSIP, a share count, and a market value. It tells you nothing about the firm's total AUM, which means you cannot contextualize any position. A $5 million holding means something entirely different for a $150 million RIA than for a $15 billion institutional manager. Without a denominator, you cannot identify overweights, underweights, or meaningful concentrations. Every analysis that requires normalized position sizing requires an external source of AUM data mapped back to each filer with precision, and that data is not in the filing.
The Accuracy Question Is Really a Processing Question
The raw 13F is as accurate as the filer made it… which is to say, mostly accurate at the position level, subject to CUSIP errors and the occasional confidential treatment gap. The bigger accuracy problem is what happens between the SEC and your team.
If a data provider is not continuously monitoring for amendments, the dataset drifts from reality every time a filer corrects an error. If security identifiers are not validated, positions get misattributed or lost. If AUM context is not layered in, the data cannot be meaningfully interpreted. If the filing cadence is not understood, teams make outreach decisions based on positions that may no longer exist.
The ten structural problems with raw 13F data compound each other. A missed amendment on top of a misidentified CUSIP on top of no AUM context produces confident-looking wrong answers, which is worse than no answer at all.
How Dakota's Holding Data Maintains Accuracy
Dakota's holding data was built specifically to solve these problems. Every 13F filing is ingested as it is released from EDGAR and processed through a seven-step enrichment workflow before it reaches your team. That includes:
- Continuous amendment monitoring. Every amended filing is picked up and applied, so the dataset always reflects the filer's most current, corrected record rather than locking in a snapshot from the original deadline.
- Security validation. Every CUSIP is validated against a live reference database. Errors are corrected. Share classes are resolved. Fund reorganizations and ticker changes are tracked and propagated through the full time series.
- AUM context layered in. Every filer is matched to a known allocator profile with AUM data from Form ADV and other sources, so every position can be interpreted in the context of the firm's total size.
- 19 asset classes and 236 sub-asset classes. Every position is tagged by asset class, sub-asset class, strategy, geography, and vehicle type — so a fund manager is not just seeing a ticker, they are seeing what that position represents in the context of a firm's broader portfolio.
- Contact layer attached. Every account is connected to confirmed, active contacts: names, roles, and outreach information. The data does not stop at the position.
The result is a dataset that starts with the same public SEC disclosures as any other source — and ends somewhere categorically more useful.
Book a demo of Dakota's holding data to see how the enrichment layer works across your target RIA universe.
