Positioning, filings, and the other side of your thesis
Institutional and insider filings read properly, a book that can be stressed against defined scenarios, and monitoring set on the things that would change your mind rather than on the things that confirm it.
Free plan. No card required.
What research tools does a small fund actually need?
Three that are hard to buy separately at a sensible price. Ownership and insider data parsed correctly, which sounds trivial and is where most cheap tools are quietly wrong. Portfolio risk analytics that align return series properly. And monitoring that runs without anyone remembering to check.
The specifics
Institutional ownership, parsed correctly
Quarterly manager filings aggregated per security rather than per row, because a filer reports the same name once per managed portfolio and reading the first row understates the position badly.
More on thisInsider transactions that carry signal
Form 4 filings filtered to open-market activity, sized in basis points of market capitalisation rather than in raw dollars, with plan sales set aside entirely.
More on thisPortfolio and risk analytics
Exposure, concentration, volatility and value at risk, with return series joined on trading session rather than on array position.
Scenario stress testing
What a defined market shock does to the book, position by position, using betas that are correct for assets trading on different calendars.
Bear case, argued explicitly
The opposing case constructed as its own argument, then scored, rather than a balanced summary that commits to nothing.
Scheduled monitoring
Workflows that watch for a filing, a price condition or a change in ownership and produce a written result on their own schedule.
The comparison in full
| Detail | Why it matters |
|---|---|
| Aggregate holdings per security, not per row | A large manager lists one name several times, once per portfolio |
| Reject principal amounts in a share count | A bond's principal summed into shares inflates a position out of nothing |
| Count managers, not dollars | One very large book would otherwise be the entire signal |
| Join return series on session date | Index alignment inverts the sign of a beta on any asset trading a different calendar |
| Distinguish a failed fetch from no activity | Otherwise an unreachable source reads as a quiet company |
Five parsing decisions that separate ownership data you can use from ownership data that looks fine.
A quarterly filing is not a current portfolio
Institutional holdings are filed 45 days after quarter end, cover long US equity positions only, and exclude shorts entirely. On the day one is published its oldest position is four and a half months old. It is genuinely useful for understanding what a set of investors has been doing. It is not a live view of anyone's book, and any tool that presents it as one is selling you a lag.
Frequently asked questions
Can I track institutional ownership changes?
Yes, across a curated set of tracked managers, comparing quarter over quarter with a materiality deadband so that rounding and corporate actions do not classify every position as changed. The set is curated rather than exhaustive, and the interface says so, because absence from it means these managers rather than the whole market.
How is insider activity scored?
Only open-market purchases and sales are counted. Option exercises, tax withholding, awards and gifts are compensation plumbing and are excluded. Sales under a pre-arranged plan are set aside rather than scored, and size is measured in basis points of market capitalisation so that a large company does not look bearish simply for being large.
Can I stress test a portfolio?
Yes, against defined scenarios. The detail that matters is that return series are joined on trading session rather than by position in an array, which sounds pedantic until you note that index alignment inverts the sign of a beta for anything trading a different calendar and can present the riskiest holding in a book as a hedge.
Can research run on a schedule?
Yes. Workflows run on a trigger or a schedule, do the retrieval and the writing, and post the result. Runs are capped per plan, and a blocked run is reported rather than silently skipped.
Check a name you already have a view on
Ownership, insiders and the bear case, on something where you will notice if it is wrong.