The standard reading of an insider purchase in a heavily shorted name is informational. Someone with a better view of the business is buying it, so the short thesis is probably wrong. That reading is available to everyone with a screen and is priced accordingly. The reading that is not automatically priced is mechanical: the purchase removes shares from the tradable and lendable pool, and in a name where the short base is large relative to that pool, a small mechanical change can be worth more than the information content.
What the overlay joins, and the date stamp that governs everything
The Squeeze view pairs two feeds and says so on the page. It describes itself as tickers where insiders are buying while short interest is elevated, with data pairing insider transactions against the bi monthly short interest report, and the footer names the sources as the bi monthly Short Interest Report plus the insider filing feed, noting that short interest is published twice per month on the fifteenth and the last business day.
Directly under the title sits the line that governs how you may use any of it. At capture it read FINRA short interest as of 2026-03-30, with 26,353 tickers covered. That is the vintage of the short leg. The insider leg is a filings feed and moves daily. So the join is between something continuous and something discrete, and the discrete side carries a settlement and publication lag on top of its fortnightly cadence.
Treat that asymmetry as the primary modelling constraint rather than a footnote. Every DTC figure on the screen has a numerator with a date on it and a denominator that has kept moving since. An insider bought after the short interest snapshot cannot be reflected in the short leg at all, which is precisely the case you are hunting.
Days to cover moves through the denominator, not the numerator
Be precise about the identity before modelling anything with it. Days to cover is shares short divided by average daily volume. An insider open market purchase does not reduce shares short. Nobody covered. What it does is act on the denominator and on the pool from which the numerator can be sustained.
The denominator effect is the one people get backwards. In the short run, an insider buy and the attention it draws typically raise average daily volume, which lowers reported days to cover even though the position is harder to exit, not easier. A falling DTC in the fortnight after a cluster of purchases is therefore ambiguous and should never be read as shorts covering without the next short interest print to confirm it. This is the single most common misreading of the metric on a desk.
The durable effect runs through float. Shares bought by an affiliate and held become control securities for resale purposes, subject to the volume and manner of sale conditions that govern affiliate sales, and they are usually held in accounts that do not participate in securities lending. That takes them out of the lendable pool in a way that ordinary retail buying does not, since retail shares in margin accounts are lendable and insider shares in a broker account generally are not being lent against a short.
Sizing the float effect before you call it a mechanic
The test is arithmetic and it disqualifies most candidates immediately. Express the purchase as a percentage of free float, then as a percentage of the shares currently on loan.
| Free float, USD | Insider purchase | Purchase as percent of float | Short base at 20 percent of float | Purchase as percent of short base |
|---|---|---|---|---|
| 5,000m | 1m | 0.02 | 1,000m | 0.1 |
| 500m | 1m | 0.20 | 100m | 1.0 |
| 120m | 1m | 0.83 | 24m | 4.2 |
| 40m | 1m | 2.50 | 8m | 12.5 |
The same one million dollar purchase is a rounding error in the first row and a meaningful bite out of lendable supply in the last. That is the whole point. A screen that ranks by SI percent of float, as this one does, will happily put both rows in front of you, and only the bottom of the table has a mechanical story. The information story might still be present in the top row, but it is a different trade with a different holding period and a different reason to be in the book.
Set a threshold in advance rather than after seeing the names. If the purchase is below one percent of the estimated short base, treat the setup as informational only and size it on the information. Above a few percent, the financing channel becomes worth modelling, and you should be looking at the trajectory of borrow rather than the level of short interest.
Borrow cost is not on this screen and belongs in the thesis anyway
The captured table header carried Ticker, Company, Insider, Filed, Trade value, SI percent float sorted descending, DTC and Signal. There is no borrow fee, no utilisation and no rebate rate. I looked for one and it was not there, so I will not tell you the module supplies it. The financing side has to be sourced from your prime broker or stock loan desk and joined on ticker, which means the squeeze thesis you can build here is half a thesis until you do that join.
Once you have it, the ordering of evidence changes. Short interest as a percentage of float is a stock measure published fortnightly with a lag. Borrow fee and utilisation are continuous prices for the same scarcity, set by people with capital at risk today, and they lead the printed short interest rather than following it. A rising fee against a stable published short interest tells you lendable supply is tightening now, which is exactly the observation the fortnightly report cannot make. If the insider purchase mechanism is real, the fee should register it before the next short interest print does.
Two financing details that decide whether the position is holdable. Rate volatility, because a fee that moves from single digits to triple digits annualised turns a positive carry thesis into a bleed regardless of direction, and the move happens fastest in exactly the small floats where the mechanic is strongest. And recall risk on the other side of the trade, which matters if any part of your structure is short or if you are lending the long into the same tightness.
What an empty screen is telling you about base rates
At capture the filters were set to a minimum short interest of 15 percent, a lookback of 30 days, insider direction Buys and a minimum trade value of 100 thousand dollars, and the result was zero candidates, with the empty state suggesting you lower the minimum short percentage or widen the lookback.
Resist that suggestion, or at least price what it costs. The available minimum short thresholds were 5, 10, 15, 20 and 30 percent, and the lookbacks were 7, 30, 60 and 90 days. Dropping from 15 percent to 5 percent does not find more of the setup you were modelling, it changes the setup, because at 5 percent of float the short base is rarely large enough for a one million dollar purchase to matter mechanically. Widening the lookback from 30 to 90 days has a subtler cost: it admits purchases that are older than the short interest snapshot itself, so the join stops being informative in the direction you care about.
The honest conclusion from an empty grid is that the mechanically interesting version of this setup is rare, which is consistent with it being worth something when it appears. Log the empty windows. A count of how many fortnights produced zero candidates at your chosen thresholds is the base rate your capacity plan needs, and it is the number that tells you whether this can be a sleeve or is only ever an opportunistic trade.