Methodology: where the data comes from and how it is calculated

Sources

• 13F: quarterly portfolios of every manager with over $100M (since 2013). What goes in and who files, per the SEC: Form 13F FAQ and the official list of 13(f) securities.
• Schedules 13D/13G: anyone crossing 5% of a company (XML since December 2024). SEC interpretations.
• Form 4: purchases and sales by each company's insiders (officers, directors and 10% owners). Form and instructions.
• XBRL: financial statements companies file with the SEC.

Every filing can be looked up on EDGAR, the SEC's public system: search filings.

We do not use market quotes: the “implied price” comes from the 13F filings themselves (value ÷ shares) and is quarterly.

Data cleaning

• Real publication dates: every data point carries the date it became public (no look-ahead).
• Splits, CUSIP changes, spin-offs, IPOs and acquisitions are detected so they don't look like buys or sells.
• Amendments, 13F-NT (portfolio reported by another entity) and filer scale errors (thousands instead of dollars) are corrected and flagged.
• When a 13F has internal inconsistencies with no single reading (values that do not match the share count, a table that adds up to more or less than the reported total), we show it as the manager filed it, with a note on its page, and that portfolio is not added to per-stock totals nor counted among concentrated managers for that quarter.
• “Traded weight”: the part of a weight change due to buying or selling, excluding price effects.

Concentrated managers (criteria 2.0.0)

Concentrated managers: 13F filers whose portfolio has 5 to 100 stocks, at least 45% in the top 10 positions and low turnover. They are investment firms, family offices and insurance holding companies, plus filers that meet those criteria and whose type we could not verify with a document.

Out of the 8,931 filers with a current 13F portfolio, we follow 703 closely and call them concentrated managers. They qualify under published rules: at least 8 quarters of history, 5 to 100 positions, 45% or more of the portfolio in the top 10, $250M or more, average turnover up to 50%, options up to 20%, ETFs up to 20%, and no single position above 70%. We also classify each filer by type and leave out those who hold stock for reasons other than a portfolio decision: operating companies with strategic stakes, banks and derivatives desks, foundations and endowments, pension and sovereign funds, venture and private equity firms whose holdings went public, credit funds and wealth advisers. Each filer gets a type. A hand-reviewed list comes first; then what the manager itself reports on its SEC Form ADV (whose money it manages and what kind of funds it runs); with no document, rules on its name and portfolio. We publish the list, the rule applied and the data point for each. Of the 703 concentrated managers, 466 have a type backed by a document, 46 have a type from the hand-reviewed list with no document behind it, and 191 are shown as “type not verified”: they meet the portfolio criteria and we found no official document that is enough to say what they are, so they may include advisers, banks or family vehicles. A type backed by a document can still be debatable: an adviser with discretionary mandates over the accounts of wealthy individuals is shown as an investment firm, with the data in view. Errors are possible. Being in this group describes the shape of a portfolio: it does not mean the manager is a better investor or has performed well.

The shape criteria, reviewed every quarter:

• at least 8 quarters of history (an entity replacing another of the same manager inherits its history);
• between 5 and 100 stock positions;
• the top 10 make up at least 45% of the portfolio;
• portfolio of at least $250M;
• average turnover up to 50% (last 4 quarters);
• options up to 20% and ETFs up to 20% of the portfolio; a security counts as an ETF by its name or by the type OpenFIGI assigns it (previously by name only, and those from lesser-known issuers slipped through);
• an adviser whose Form ADV reports 70% or more of its assets in individuals' accounts, with no other signal, is shown as “type not verified” rather than as an investment firm (it still counts: two signals are needed to exclude it);
• no single position above 70% (4-quarter average).

Turnover is the sum of the weight changes of all positions between two quarters, divided by two: it includes the effect of price moves, not only buying and selling. It is not calculated for the quarter in which a manager changes filing entity.

Filer type

Every filer has a type. It is assigned, in this order: by the hand-reviewed list; by what the manager itself reports on its SEC Form ADV (or, in a few cases, in other EDGAR filings), when that document is enough to tell; and, with no document, by automatic rules (the filer is also a listed operating company; nearly all of its portfolio is holdings that went public and it has 12 positions or fewer; words in the name such as “Foundation”, “Pension”, “Ventures” or “Bank”; an individual with more than 40 positions). If none of that settles it, it stays “type not verified”. The type, and the document backing it when there is one, are shown on each manager's page.

Download the hand-reviewed list (CSV) CIK, name, type, whether it is included or excluded, and why.

Download the rule applied and the data point for the filers reviewed (CSV) CIK, name, type, confidence, rule, source, date and the data point from the document. It does not list every filer in the table: only those that went through the document review (candidates for concentrated manager and those with a page); every filer's type is in the full table.

Download the type of every current filer (CSV) One row per filer: CIK, name, type, whether it counts and whether it is a concentrated manager, where the type comes from (reviewed list, document or automatic rule), the rule, the CRD when matched, and the note.

The lists on the site (season, most added, new positions, consensus, Ideas) count concentrated managers only. Well-known investors who do not meet the criteria keep their page and are left out of those counts.

How many managers hold a stock and how that changes

One count across the site: managers now − managers before = those who opened − those who exited + “no comparable data”. That last line is managers whose previous-quarter portfolio cannot be compared (first 13F, or they did not file), spin-offs and corporate exits (mergers, delistings): they move the count without being a clean buy or sale, and are shown separately.

“Before” is the previous quarter counted with today's set of concentrated managers. The historical chart on each page uses each quarter's own set, so its second-to-last point may differ by a few managers: the page says so when it happens.

“Most added” lists are ordered by number of managers, not dollars or shares. “Accumulation” is different: consecutive quarters in which those still holding the stock own, combined, more shares; it can happen even if most trimmed, which is why we show how many added and how many trimmed.

When a company has several share classes (Alphabet: GOOGL and GOOG), the manager page adds them up per company and the stock page shows the weight of that class and, next to it, the combined weight.

A change of less than 1% in the share count counts as “no change”: it is not counted among those that added or trimmed. And if a manager drops one class of an issuer and shows up with another class of the same issuer (same first 6 digits of the CUSIP, for example a common and a preferred share), it is shown as a class change, not as a sale and a purchase.

Insider buying (Form 4)

• Open-market purchases only (code P), excluding 10b5-1 plans, of $25,000 or more.
• Purchases in the company's IPO (at the offering price, from the underwriters or in a SPAC's private placement) are shown with their label and are not part of the totals: one is “in the IPO” if, on the same day, the same Form 4 has a footnote naming the initial public offering (unless that purchase's own footnote says it was on the open market or at an average price), or if it is from the same company, at the same price and within 3 days of one already flagged.
• One buyer = one group of filers: filings for the same issuer that share any reporting person count as one, and the amount once. A Form 4/A replaces the original.
• Officers and directors are counted as such. 10% owners with no role go in a separate block (purchases of $1M or more in 30 days) and are not part of “N officers and directors bought”.
• Relevant purchase: $100,000 or more, or 10% or more of prior direct holdings. Notable: $1M or more, or 25% or more. Group buying: 3 or more buyers with a role in 30 days adding up to $250,000 or more (2 is enough if one is the CEO or CFO).
• Two different filers with the same date, price and quantity, both indirect holdings, are counted once (“possibly the same transaction”).

Position history and quarter-end prices

• Each quarter's shares are restated on today's basis: a split does not look like a purchase. The series continues when the stock changes CUSIP and when the portfolio moves to another entity of the same manager.
• “Since”: first quarter of the current run, unbroken until today. If our history begins with the position already open, we say “since at least”.
• “It bought in N quarters”: quarters in which the share count rose by more than 1%. For those quarters we show the lowest and highest quarter-end price (reported value ÷ shares in the 13F).
• What it does not say. It is not the price paid: it may have bought at any price within each quarter, and the SEC only publishes the value at quarter end. That is why we do not compute an average cost or a gain per position, and you cannot sort or filter by this figure.
• A quarter in which the manager filed and the stock is not there shows as “no position”; one with no 13F from the manager, as “did not file”. Full history for managers that were ever concentrated; 8 quarters for the rest.

Holders among all 13F filers and institutional ownership

• For each stock we publish the 25 13F filers with the most shares and the 15 that added and cut the most in the quarter, with their filer type. The change is measured in shares, split-adjusted; if the filer did not file the previous quarter it is left empty and counts as neither a buy nor a sale.
• “Index firms” is a filer type assigned by hand to houses whose 13F is mostly index funds; the list is in the types file we publish.
• Institutional ownership = shares of all 13F filers ÷ shares outstanding on the cover of the nearest earlier 10-K or 10-Q. It is not published for ADRs, companies with more than one share class, covers older than 200 days, when there is a split between the cover and quarter end, or if the result falls outside 1% to 150%.
• What it does not say. It is neither all institutional ownership nor the shareholder list: only managers above $100M file a 13F, it excludes short positions, and entities of the same group file separately (there may be double counting and figures above 100%).

Options (puts and calls)

• Per manager, the puts and calls of its last 12 quarters with their status: “new” (not there the previous quarter), “still held” or “no longer there”. Market makers are excluded.
• The figure is the value of the underlying shares (notional), not the premium paid. It is not added to the portfolio.
• What it does not say. The 13F does not report whether the option is bought or written, or its strike or expiry. A put may hedge shares the manager owns. “No longer there” does not distinguish expiry, closing or exercise.

Stakes above 5% over time (13D/13G)

• Each holder with its filings in order, the percentage in each and the difference from the previous one. History since December 2024, when the SEC moved these forms to a machine-readable format.
• “Moved from 13G to 13D”: it had been filing the passive-investor form and now files the one for those who may seek influence. What it intends is in the filing, which we link; we do not interpret it.
• Change of filing entity: when a house reorganises who files, the old entity reports 0% and the new one appears with the stake. We join them in a single line: it is not a sale.
• What it does not say. An annual 13G/A may show months later a sale that happened earlier. Percentages are those each holder reports, over the share count the holder itself uses.

Flagged material events (8-K)

• We flag ten 8-K items. Serious: 4.02 (non-reliance on financial statements), 1.03 (bankruptcy) and 3.01 (delisting notice). Medium: 4.01 (change of auditor), 2.04 (debt acceleration), 2.06 (impairment), 1.05 (cybersecurity incident) and 2.05 (restructuring). Routine: 3.02 (unregistered equity sale) and 5.02 (change in officers or directors).
• Severity belongs to the item type, the same for every company. We show the date, the item, the original press-release headline when there is one (not summarised) and the link to the filing.
• What it does not say. We do not read the content: Item 5.02 mixes routine appointments with a chief executive's departure, and a 3.01 may be a notice cured within weeks. No flags does not mean no problems. It covers recent months, not the full history.

Company fundamentals

• Source: the financial statements the company files with the SEC (10-K and 10-Q) in XBRL, under US accounting standards (US GAAP) and in dollars. Foreign issuers that file under IFRS (20-F) or in another currency, and ETFs, have no fundamentals: we do not mix standards or currencies.
• 12-month figures: those of the latest 12-month period filed; the page says when they end. Operating margin = operating income ÷ revenue. Net margin = net income ÷ revenue. ROE = net income ÷ equity on the latest balance sheet. Revenue and earnings growth = 12 months versus the prior 12 (earnings growth is left empty if there was a loss before).
• Free cash flow = operating cash flow − capital expenditures (capex). If the company does not report capex, free cash flow is left empty: we do not take it as zero. Free-cash-flow margin = free cash flow ÷ revenue.
• Cash = cash and cash equivalents on the balance sheet. Investments = current and non-current marketable securities; stakes in other companies and non-marketable investments are not included. Total debt = short-term borrowings + bonds (current and non-current portions) + finance leases; operating leases are not included. If the company does not report a total, we add up each instrument and the page says it is approximate. No debt data is not “no debt”: we do not calculate net debt or its ratios.
• Net debt = total debt − cash − investments (negative = net cash). Net debt / EBITDA uses 12-month EBITDA = operating income + depreciation and amortization; if the company does not report depreciation and amortization there is no EBITDA, and the ratio is also left empty if EBITDA is not positive. Cash, investments, debt and equity come from the same balance sheet, whose date is shown; if that balance sheet is more than 15 months older than the 12-month figures, we do not show them.
• Banks, insurers and other financial companies (by SIC code: 6000–6199, 6200–6299 except 6282, and 6300–6399 except 6324, health insurers): we do not show free cash flow, cash, debt or their ratios. Their operating cash flow moves with loans and the investment portfolio, and their debt is part of the business (deposits, funding). REITs are not in this group.
• By fiscal year: each bar is one fiscal year, labelled with the year the company reports for it (if missing, the year it ends; a year end in the first 7 days of January counts as the previous year).
• Impossible ratios are left empty: margins above 150%, ROE above 300%, revenue growth above 1,000% or earnings growth above 5,000% (usually tags that are not the total, in banks and REITs, or misreported scales).
• “Quality vs sector”, 0 to 100: the average of the company's percentiles in ROE, operating margin, free-cash-flow margin, revenue and earnings growth and net debt / EBITDA (for debt, the least indebted is best; at least 3 needed), against companies in its sector with current statements and a market value above $300M (or, without it, revenue above $1,000M). It is not shown if the sector comes from the unreviewed SIC code or has fewer than 20 comparable companies.
• What it does not say. These are mechanical calculations on what each company tagged in its statements: they do not replace reading the annual report. They are not a recommendation or a judgement on the company.

Buybacks, dilution, earnings quality and debt

• Shares outstanding: average diluted shares of each fiscal year (10-K), restated on today's basis with known splits. The 1, 3 and 5-year change is left empty if the window contains a jump of more than double or less than half with no known split: it may be real dilution or an undetected split, and we would rather not assert it.
• Buybacks, issuance, stock-based compensation and dividends: from the latest fiscal year's cash flow statement. Not every company reports each item under a readable tag: “no data” is not zero.
• Cash flow over net income = operating cash flow ÷ net income, 12 months (not calculated with negative net income). We warn when more than 25% of net income does not come from operations.
• Interest coverage = operating income ÷ interest of the latest fiscal year. Debt due within 12 months, per the balance sheet, and against cash. Maturities by year, from the debt note of the annual report. For banks and insurers these ratios do not apply and are not shown.
• What it does not say. These are mechanical calculations on what each company tagged in its statements: we do not check them one by one against the annual report and they may contain errors. They are not a judgement on anyone's solvency.

Each manager over time and the two turnovers

• Per quarter: positions, reported value, weight of the top 10 and of the largest, and the sector split. A new entity's series includes that of the manager's previous entity.
• Turnover: the sum of the weight changes of all positions, divided by two. It includes the effect of price moves, and is the one used by the concentrated-manager criteria.
• Traded turnover: only the weight the manager bought and sold, excluding price. It is lower and better describes how much it traded. Both are left empty in the quarter of the entity change.
• What it does not say. Reported value moves with prices and with the manager's client inflows or outflows: it is not its performance.

Sectors

Each company's sector comes from the industry code (SIC) it reports to the SEC, mapped to eleven sectors with the table below. SIC is a 1937 classification and does not always match others: that is why we correct by hand a short list of widely held companies (for example, Alphabet and Meta go in Communication Services; Visa and Mastercard, in Financials). A sector's weight is the average, across concentrated managers, of its weight in each one's portfolio: every manager counts the same. What we could not classify is reported separately as “no sector”.

We also reviewed, one by one, the companies with the most value in concentrated managers' portfolios (about 1,400) and assigned their sector by business, not by code: that list covers the great majority of the value held. Each stock page says where its sector comes from: reviewed or “by SIC code, unreviewed”. Where it is by SIC, the “versus its sector” comparison may not be representative and the page says so. The review is our own (market classifications such as GICS are licensed) and may contain stray errors.

Consensus score (consenso-v2)

0 to 100, only for stocks with 5 or more concentrated managers (not calculated below that). It is always shown next to the number of managers:

• Accumulation 40%: consecutive quarters in which its holders own, combined, more shares (capped at 4).
• Conviction 30%: percentile of the median weight in its holders' portfolios.
• Net managers 20%: percentile of entries minus exits in the quarter.
• Breadth 10%: percentile of the number of managers.
• −10 if no manager added in the quarter.

What it does not measure. It does not measure whether more managers are buying than selling or how many dollars moved, and it does not mean managers “agree”. A high score can sit alongside more managers trimming than adding (shown next to it, on each stock) and very few holders (5 are enough).

What our study says. Our own study supports only one of the four factors: accumulation. Conviction, net managers and breadth, which add up to 60% of the score, showed no reliable edge; they are there because they describe the holding, not because they anticipate returns. The weights were chosen looking at the study's same historical sample. It is an order to start researching, not a signal. There is no test on later data yet.

Net flow in shares and dollars

Next to the count of managers we show, per stock and quarter, the flow among concentrated managers: the shares added by those who opened or increased minus the shares cut by those who trimmed or exited. In dollars it is valued at the 13F quarter-end implied price (the median, across filers, of reported value divided by shares; not calculated with fewer than three filers). It is not the price traded at: the 13F does not say. Rows with no comparable quarter, spin-offs and corporate exits do not count.

The flow does not change the order of any list or enter the consensus score: it is a second measure. A stock can gain managers and have a negative flow (many enter with little while a few sell a lot), and the other way round.

Downloads: the full data by quarter

Concentrated managers' holdings (with each 13F's address), changes by stock with their flow, insider transactions and stakes above 5%, quarter by quarter, as compressed CSV and Parquet.

The download files and the study (with its tables, charts and observations) are published under Creative Commons Attribution 4.0 International (CC BY 4.0): you may use, copy, adapt and redistribute them, including commercially, with attribution (“GestorLens (gestorlens.com), with SEC data”). The original SEC documents are in the public domain. The rest of the site, the GestorLens brand and the design are not under that licence. This is not investment advice. CC BY 4.0 · Data licence.

Observations of the study “Does copying superinvestors work?” (one row per stock and quarter): Parquet · CSV · column dictionary and how to verify · the study

“If you had copied it” and manager ranking

Each 13F portfolio is bought at the first quarter end after it was published (what anyone could have copied), at the reported weights, and held for one quarter. Only what was known at the first filing counts: confidential positions revealed later could not have been copied.

It is compared against SPY over the same dates. Prices implied by 13F filings (at least 3 filers per price), split-adjusted. It is price return, with no dividends or costs, for SPY too: that is why the index figure is lower than its total return, and high-dividend portfolios are somewhat penalised. The ranking includes managers no longer counted as concentrated, to avoid survivorship bias.

It measures copying the 13F, not the manager: for a global firm the 13F is only the US-listed slice. It is an ordering by past results and does not say who the better investor is.

What our studies show

129,534 observations of 8,228 stocks (one observation is one stock held by concentrated managers in one quarter) across 48 quarters (2013–2025), with 12-month price returns versus the S&P 500 and Newey-West t-statistics. It measures the simple average of those stocks, not copying a portfolio with its weights:

• Most 13F signals, on their own, do not give a statistically reliable edge.
• The most consistent signal was accumulation: stocks nobody was adding to did worse than the whole set (t = −3.60) and the gap in favour grows with the quarters of accumulation (from +0.5 pp to +2.2 pp), although the 4+ group does not reach the usual threshold (t = 1.88). High conviction and company profitability did not give a statistically reliable edge.
• That is why we sell clarity, time and verifiable data, not returns.

Read the full study: Does copying superinvestors work? →

Limitations and conflicts of interest

• 13F filings come out up to 45 days after the quarter and only show long U.S. equity positions: portfolios may have changed.
• 13F filings only cover section 13(f) securities: US-listed stocks, ADRs, ETFs and some options and convertibles. Foreign shares, bonds, cash, short positions and OTC derivatives are left out. They do not say whether an option is bought or written.
• Data may contain errors or omissions. What can be audited in one click: each manager page links to the exact 13F document for the quarter (and its amendment, if any), and each 13D and 13G to its document. Flagged 8-K items and each step of the 13D/13G history link to their filing. Form 4: every trade in the latest-trades table (7 days) links to its filing; the 30-day per-company purchase summary and the one on each stock page are totals and link to the company's Form 4 filings on EDGAR, not to each one. Each position shows its CUSIP and links to the 13F where it is reported (the document, not the row: a 13F is a single table). Financial-statement figures do not have their own link yet.
• The service provider may hold positions in securities shown on the site. Categories and rankings come from mechanical rules applied equally to all.

Glossary of terms

Criteria change log

• 2.0.0 · Oct 9, 2026. The group we follow is now called “concentrated managers” and gains a filer type: operating companies, banks and desks, foundations and endowments, pension and sovereign funds, private equity and venture capital, credit funds and wealth advisers are left out. New criteria: no position above 70% and ETFs up to 20% (previously 25%). An entity replacing another inherits its history. The group went from 858 to 714 managers as of 30 June 2026, and the counts, the consensus (now from 5 managers), the ranking and the study changed with it. With the evidence-based classification (Form ADV and EDGAR) of the same day, the current figure is the one in the table above.
• 1.1.0. Shape criteria (history, positions, concentration, size, turnover, options and ETFs) and a name-based exclusion for banks and insurers.

Corrections log