Methodology

The filing facts, the computed figures, and the rules between them

The filings behind this site are public, so the useful distinction is how carefully they are read. This page states what is sourced, what date it refers to, what is computed rather than filed, how the stored 13F excerpt is capped, and what is deliberately left out.

The two windows the overlap uses — 120 days on the 13F leg and 45 days on the Form 4 leg — are defined on the overlap page.

Disclosures shown are public regulatory filings. Data may be delayed or incomplete. Smart-money entities may no longer hold positions shown. Not investment advice.

Provenance

Where the data comes from

  • Form 13F-HR and 13F-HR/A information tables, as filed with the SEC and published on EDGAR. Flock reads the filing itself, not a third-party summary of it.
  • Form 4 and Form 4/A statements of changes in beneficial ownership, from the same source.
  • No broker feeds, scraped estimates or analyst inputs feed the source data. Quarter deltas and overlap strength are Flock computations derived from filed figures under the rules below.

Timing

Quarter-end is not filing date

Almost every misreading of 13F data is a date problem, so the dates are stated first.

  • A 13F position is as of the QUARTER-END it describes, not the day it was published. A 13F-HR may be filed up to 45 days after that quarter-end, so the newest 13F position on this site is already at least weeks old and may be considerably older.
  • A Form 4 describes a dated transaction and is due within 2 business days of it. The transaction date says when it happened; the filing date says when it became public.
  • Flock stores both dates when the SEC supplies them. The 45-day overlap window and its filing timeline use the filing date, because the event cannot be observed before the disclosure is public.
  • No figure on this site is a current position. A filer may have bought or sold the whole position the day after the period it reported, and would not have to say so until the next filing.

Comparisons

How a quarter-over-quarter change is computed

  • A 13F reports an absolute snapshot, never a change. Every quarter-over-quarter figure Flock shows is therefore computed by Flock, by comparing two snapshots.
  • The comparison is made against a NAMED prior period that is recorded with the result, so any delta can be checked by opening two filings on EDGAR.
  • Where no defensible prior snapshot exists — the filer is new to Flock, skipped a quarter, or the prior filing is incomplete — no delta is computed and none is shown. An absent comparison is left absent rather than inferred from silence, because a missing filing is not evidence of a zero position.
  • The current dataset retains up to the 200 largest positions by reported value per tracked filer-quarter. A prior quarter at that cap is treated as incomplete for absence-based comparisons.
  • If a security is absent from a stored prior filing that is below the cap, that absence is treated as a defensible zero. Otherwise a first appearance is shown without an inferred increase.

Ordering

How overlap strength is computed

  • Overlap strength is a 0–100 display and ordering score. It does not decide whether an event qualifies, and it is not a rating, recommendation or estimate of future performance.
  • Breadth is 1 ÷ (1 + e^(−0.8 × (party count − 2))). Party count is the number of qualifying 13F filers plus distinct Form 4 reporting owners.
  • Recency is computed separately for the 120-day 13F leg and the 45-day Form 4 leg as the mean of e^(−age ÷ window), then the two leg values are averaged. Age is measured in days to the moment of computation, from the quarter end a 13F reports (its period of report) and from the filing date of a Form 4 — the two legs are dated from different things because a 13F describes a quarter already past while a Form 4 is dated when it is filed.
  • Final strength is 100 × breadth × (0.7 + 0.3 × recency), capped at 100 and rounded to two decimal places.

Form 4

What counts as a purchase disclosure

  • Only Form 4 rows with SEC transaction code P count as a Form 4 purchase disclosure. Awards (A), gifts (G), option exercises (M), sales (S) and dispositions are excluded — they are different acts and reading them as buying would be wrong.
  • SEC transaction code P records a purchase without stating where it was executed, so Flock describes these filings as purchase disclosures and nothing more.
  • The reporting owner's relationship to the issuer — officer, director, ten-percent holder — is taken from the filing and shown as filed.

Corrections

Amendments, supersedes and exclusions

  • An amended filing supersedes the row it corrects. The corrected row is retired, not left standing alongside the correction, and anything Flock derived from it is retired with it.
  • Where an amendment cannot be matched confidently to the row it amends, neither row is used for a comparison. An ambiguous correction is treated as a reason to say nothing, not as a reason to pick one.
  • A superseded filing can retire an overlap that was already raised. When that happens the overlap is withdrawn rather than quietly left in place.

Identity

Matching a 13F holding to a Form 4 issuer

  • A 13F identifies a holding by CUSIP; a Form 4 identifies its subject by issuer CIK. Joining the two is a mapping step, and it is Flock's mapping, not the SEC's.
  • Where the mapping is not confident, the two filings are not joined and no overlap is raised. A missed overlap is a cost Flock accepts; a wrong one is not.
  • Multi-class issuers are the known hard case. A Form 4 names the issuer and the class of security in its own terms, which do not always resolve to the exact share class a 13F reported. Treat a multi-class name with more caution than a single-class one.
  • Ticker changes, mergers and issuer renames are handled at the issuer level, so a filer's history does not silently split in two when a company changes its name.

History

Backfills never notify

  • Historical loads run with notifications suppressed. Importing or re-importing years of filings never emails anyone — an alert about a 2019 quarter would be noise, and receiving one would rightly destroy trust in every other alert.
  • A backfilled overlap is still recorded and still visible on the site. It is the notification that is suppressed, not the fact.
  • Re-processing a period after a parser fix can change what is shown for that period. That is the intended behaviour: the site reflects the filings as currently read, and a correction is applied rather than preserved for consistency's sake.

If something here looks wrong

Every SEC source row names the filing it came from, so a disagreement can be settled by opening it. If a figure does not match the filing, that is a defect and we want to know — the data enquiry form reaches a human.

Flock republishes information from public SEC filings for informational purposes only. Nothing on this page is a recommendation, solicitation, or offer to buy or sell any security.

Flock is not a registered investment adviser or broker-dealer, does not provide personalized financial advice, and owes no fiduciary duty to any reader.

Filed data lags reality: Form 13F holdings are as of quarter end and filed up to 45 days later, and Form 4 insider transactions are filed within 2 business days of the trade. Positions shown may have changed or closed, and filings may contain errors we do not correct.

Flock makes no warranty as to the accuracy, completeness, or timeliness of any data shown and accepts no liability for decisions made in reliance on it. Consult a licensed financial adviser before making investment decisions.

The Smart Money Digest

A free weekly email of notable disclosure activity — every line with its filing date and source link. No advice, just filings. Unsubscribe anytime.