Market Memos · Where To Look
Money Flows
Filed transactions, not opinions — who is actually buying, and how stale it is by the time you see it.
Everything on this page is a trade somebody actually made and was legally obliged to disclose. That makes it the highest-weight material available to a retail investor: these people paid for being wrong, and they filed because the law said so, not because they wanted you to know.
It is also the only category where the main risk isn't being deceived. It's being late.
How stale is it when you see it?
Read this before anything else on the page. It decides what each source can be used for.
| Filing | Lag | What it's good for |
|---|---|---|
| Form 4 insider buy | 2 business days | Near live. The best free signal there is. |
| Schedule 13D activist stake | 5 business days | Fast. Somebody intends to make something happen. |
| Form 8-K, Item 5.02 management change | 4 business days | Fast, and it changes the story rather than confirming it. |
| Congress STOCK Act | up to 45 days | A reason to look at a name, never to time one. |
| Form 13F fund holdings | up to 135 days | History. Research a name you have; never find one. |
13Fs are due 45 days after quarter end, so a position opened on 2 January first appears in mid-May and may already be closed. Anyone treating a new 13F position as news is reading about last season.
Insider buying
Unusual options activity is a genuine version of this signal, and useless on small companies — most have no real options market, so there is no volume for anything to be unusual against. The small-cap substitute is insider buying.
OpenInsider's cluster buys page is free and is the one that matters: several insiders at the same company buying in the same window. One officer buying is a data point. Three in a fortnight is a conversation that happened inside that building.
- Open-market purchases only — transaction code
P. Exercises and grants are pay, not conviction. - Officers and directors, not ten-percent owners. A fund crossing a threshold is doing mechanics.
- Size relative to the person. $40,000 from a director beats $2m from a founder holding nine figures.
- Ignore selling. Most of it is scheduled 10b5-1 plans, tax withholding or gifts. Buying has one motive; selling has twenty.
For buys and sells in one table, Finviz is the cleanest free view; SecForm4 and GuruFocus cover the same filings. They differ in presentation, not in what they know.
One caveat on OpenInsider itself: it is run anonymously, with no company behind it and no API. Reliable for years, but there is nobody to complain to.
Activist stakes
A 13D is filed within days of crossing 5% with intent to influence — somebody announcing they mean to make something happen. A 13G is the passive version and matters far less. At small-cap size a 13D can be the whole story.
Free on EDGAR full-text search — form types SC 13D and SC 13G. Fintel packages them more readably; use the free pages and read independent reviews before putting a card in.
Most reported daily short volume is routine broker mechanics — a market maker selling to you marks the trade short although no bet against the company exists. One documented case: 140 million shares of cumulative daily short volume over a fortnight against a real short-interest rise of 1.5 million.
Genuine short interest is reported to the exchanges about twice a month, on a lag. Anyone quoting a daily figure as evidence of a squeeze is quoting the wrong number.
Congressional trades
Capitol Trades is free by policy and run by 2iQ Research, a Frankfurt firm with twenty years in insider-transaction data — the best-backed operator in this category. Quiver Quantitative has a substantial free tier, and Unusual Whales covers the same ground.
All three inherit the same ceiling: up to 45 days of disclosure lag, amounts given only as wide ranges, and late or amended filings that are common rather than exceptional. Useful for noticing that a name or a sector keeps recurring. Useless for timing.
Institutional holdings — 13F
WhaleWisdom is free for the trailing two years, which is enough; full history and exports run $300–500 a year. It earns some trust by admitting in its own FAQ that filings may be parsed incorrectly and telling users to verify against SEC records — a warning that generalises to every scraper in this space.
Dataroma tracks a curated set of well-known managers and is free. EDGAR has the raw filings.
Check the period label on any holdings page before believing it. As of 19 September 2026, Dataroma was showing Q1 2026 — Q2 2026, due in mid-August, had not appeared, and nothing on the site said so. One quarter behind is survivable. Not knowing which quarter you are reading is not.
Beyond the lag, 13Fs are long-only and exclude short positions, so the picture is half a picture.
New management
The odd one out here, because it isn't a trade. A new chief executive at a small company doesn't confirm a thesis, it replaces one — everything a screen knows about that company describes the previous management's decisions.
The filing is Form 8-K, Item 5.02, within four business days of any CEO, CFO or board change. Free on EDGAR — filter to form type 8-K and read Item 5.02. Boardroom Alpha packages the same filings with a board-quality rating, but shows only a handful publicly; the tracker is a seven-day trial then paid.
The pairing worth watching for: a new chief executive who then buys stock with their own money. An 8-K Item 5.02, then a Form 4 a few weeks later. Both free filings.
An incoming CEO has the clearest possible view of what they have just walked into, and no obligation to buy. When they do, inside their first months, they are telling you something no screen can see.
The filing nobody checks
All of the above is people buying. This is the company selling — and it is the mechanism that turns a correct thesis into a losing position. A small company whose story is working will often sell stock into the enthusiasm. The company grows and your slice shrinks.
- An
S-3shelf registration is permission to issue, filed before it is used. Entirely normal — so its presence is information, not an accusation. - An at-the-market agreement is worse, because it is continuous. A company with a live ATM can sell into your rally day after day with no announcement until the next quarterly report. Search its filings for the phrase at-the-market.
- A
424B5is the offering itself. One appearing the morning after good news is the good news being sold. - Share count over time. Rising share count against flat revenue is the whole warning in one line.
All free on EDGAR full-text search, which covers filings since 2001 and can be filtered by company, form type and date. Check for a live shelf before buying a catalyst, not after.
This page covers Tier 1 — money at risk, disclosure required. For the earlier and less reliable signals, see Sentiment Reads. For why any of it is worth the trouble, The Ten-Bagger Arithmetic. Everything else is indexed on Where To Look.
