- FINRA Rule 4560 requires broker-dealers to report short positions twice a month, submit them by 6 p.m. Eastern on the second business day after the reporting settlement date, and FINRA publishes the compiled figure on the seventh business day after that same settlement date.
- The count covers positions held at FINRA member firms, which means short exposure taken through swaps or options sits outside it. FINRA declared comments asking it to disclose synthetic positions "outside the scope" of its May 2026 rule filing.
- SR-FINRA-2026-012 would move publication to weekly, five business days after the settlement date. That closes the timing gap and leaves the coverage gap exactly where it was.
A short interest figure carries two dates, and the one that decides what it means is not printed on it.
The visible date is the day the number reached you. The other one is the day the position was counted, and the distance between them is set by a rule rather than by anyone's delay. Take a real line from FINRA's published calendar: positions as they stood at the close on 14 November, submitted by firms on 18 November, released on 25 November. Eleven calendar days, during which the number sat finished and private while the market kept trading against it. Nothing in that sequence went wrong. That is the sequence working exactly as written.
That is why a headline carrying more than 116 million shares of a software company sold short, worth over $2.1 billion and around 29% of that company's float, is describing a day that has already closed by the time it reaches a screen. The figures are accurate. The settlement date they belong to is the part that never travels with them.
Eunha reads that headline the way she reads all of them, which is backwards. Kodex's interpreter goes to FINRA's publication calendar first, finds the settlement date the figure belongs to, and only then looks at the 116 million shares. "Read the date before you read the size," she says. "Only one of them is about today, and it is not the size."
Timing is the smaller of her two problems, though, and the larger one is already sitting in the definition.
What short interest counts, and what it counts against
Short interest is the number of shares sold short and not yet bought back, reported by broker-dealers for the customer and proprietary accounts on their books. FINRA collects those reports for every equity security and publishes the aggregate. The regulator is unusually direct about what the result is: FINRA's investor page calls short interest "just a snapshot that reflects short positions held...at a specific moment in time."
Eunha leaves that sentence on the screen, because it is doing two jobs at once and only the first one gets read. It tells you the figure is a still image. It also tells you the still image is of positions, not of activity, and those come apart faster than you would expect. FINRA states plainly that short interest position data does not "equate to the daily short sale volume data." Volume counts the shorting that happened. Short interest counts the shorting that is still open. A stock can print heavy short-sale volume all fortnight and show a flat short interest figure, because the same shares went out and came back inside the window between two snapshots.
There is a second choice buried in every percentage, and it sits underneath the line rather than above it. Float and shares outstanding are both legitimate denominators, and the same position measured against each produces two different percentages, both honest. Those mechanics belong to IPO lockup expiry, which works them through properly on a live example. What Eunha wants from it here is only the habit: when a percentage arrives, find out what is underneath the line before you carry the number anywhere.
The three dates between the position and your screen
Eunha puts the three dates in order, because the order is where the lag is manufactured. That count happens on a settlement date, and FINRA designates two of them a month: one around the middle, one at the end. Firms report their positions as of the close of business on those days, and everything after that is transmission.
FINRA's filing calendar fixes the rest of the chain. Members must submit their short interest positions by 6 p.m. Eastern on the second business day after the designated reporting settlement date. FINRA then compiles the reports by security and provides them for publication on the seventh business day after that same settlement date. Three dates, two of which you will never see: the day the position existed, the day the firm filed it, and the day it reached you.
Seven business days is the rule, and it is not what you experience. Business days and calendar days diverge, weekends land where they land, and a holiday inside the window pushes everything right. That November line runs eleven calendar days from count to publication. A window containing Thanksgiving or the turn of the year runs longer. The rule is fixed and the felt lag is not. So the same regime gets described as "about a week" by the people who wrote it, and as closer to a fortnight by the people waiting on it, and both are describing the same calendar. One counts business days. The other counts the days you live through.
One underlying event producing more than one legitimate published number is a shape that recurs across markets. Price-weighted vs cap-weighted indices is the same problem wearing a different hat.
None of this makes the figure wrong. It makes the figure a description of a moment you cannot act in, which is a different property, and one that no amount of care in reading fixes.
What does short interest miss when the number is fresh?
Suppose the lag disappeared. Suppose the number reached you the same afternoon the position was counted. Eunha's question is what you would be holding then, and the answer is the part of the problem that no change to the calendar reaches.
Rule 4560 reaches the short positions held in customer and proprietary accounts at FINRA member firms. That is a perimeter, and perimeters have an outside. Short exposure can be assembled through a swap, or through the delta on an options position, and neither one produces a reported short position at a member firm. The economic bet is short. The regulatory count is not.
FINRA says as much in its own filing. In the rule change it sent to the SEC in May 2026, the regulator acknowledges that "it is possible for customers to obtain short exposure to a security through other means." When commenters asked it to go further and require disclosure of synthetic long positions, FINRA recorded the request and set it aside as "outside the scope of the instant proposed rule change and therefore are not addressed herein."
Then there is the industry's own account of what a reporting boundary does to the thing it encloses. FINRA proposed to widen the count to include customer stock borrows in member firms' arranged financing programs. Fidelity objected on the record. That change would shift borrowing away from those programs, it warned, and toward "swaps dealers, custody banks, or off-shore entities that are not subject to similar reporting requirements." FINRA kept the expansion in. Those positions already sit on the member's books, it argued, and counting them would "better reflect the actual short sentiment in an equity security."
Read the objection carefully, because it is not an accusation and it should not be turned into one. Nobody in that exchange is describing concealment. Reporting rules bind the firms a regulator can compel, and the activity outside that boundary is legal, disclosed elsewhere or not at all, and structurally invisible to this particular count. Neither side argued about whether the outside exists. Both took that as given and argued about which side of the line the borrowing would end up on.
A disclosure regime can only see what it has the authority to demand, and FINRA's authority runs to its members. The perimeter of the number is the perimeter of the regulator's reach. It was never drawn around the risk, and it was never claimed to be.
So the figure is not merely late. It is late and partial, and only one of those two things is a timing problem.
Weekly reporting, five business days out
FINRA has written down the fix for the timing half, which is the strongest available evidence that the timing half is real.
On 1 May 2026 the regulator filed SR-FINRA-2026-012 with the SEC. The Commission's notice published on 18 May, comments closed on 8 June, and on 6 July the SEC designated a longer period for its decision, setting 14 August as the date by which it would approve, disapprove, or begin proceedings. What happened on or after that date is not established in the public sources available at the time of writing, so treat the proposal as a filed proposal and nothing more.
Inside that filing, FINRA proposed to require members to submit short interest reports weekly rather than twice a month, and to cut the reporting turnaround from two business days to one. Together, in FINRA's own description, those changes would let short interest data be "published weekly, five business days after the reporting settlement date." Alongside it, new Rule 4321 would require members to report monthly their daily allocations of fail to deliver positions to correspondent firms. That tells the regulator which firm actually owns a close-out obligation. Today it has to ask after the fact.
Seven business days becomes five. Twice a month becomes weekly. A smaller gap closes too, one that lives entirely inside the plumbing. When a security loses its ticker symbol before a settlement date, firms currently drop it from the report. The last short interest in a delisted or cancelled security has been disappearing that way for years. FINRA proposed a final report using the last settlement date for which a symbol was in effect.
Every one of those changes is a cadence change. Not one of them touches the swap, the option, or the offshore entity. The regulator proposed to make the photograph more frequent and left the frame exactly where it was, and it said so on the record when it declined the synthetic disclosure request. If the rule is approved in the form filed, the number will arrive faster and it will still not count what it does not count.
Crypto shows the book. Equities show a filing.
This is where an instinct you brought from crypto misfires, and Eunha thinks it is worth naming rather than correcting.
In perpetual futures, aggregate positioning is a market data product. Open interest updates continuously, refreshes at every funding interval, and sits on the same screen as the price. How to read open interest in crypto works through what that figure does and does not tell you, and the argument there stands on its own. What matters here is only the shape of the thing. The venue publishes positioning as a feature, because the venue holds the positions. Showing them costs it nothing.
In US equities, positioning is a regulatory disclosure. Nobody holds all of it. It is assembled from filings by hundreds of firms, compiled by a self-regulatory organisation, and released on a calendar published a year in advance. That is not a worse system. It is a different answer to a different question, produced by a market where the positions live in a thousand places and someone had to be compelled to add them up.
The failure mode is importing the reflex. A number that updates continuously invites you to treat its latest value as the current state of the world, and that reflex is correct where it was learned. Carried across, it turns a fortnight-old regulatory filing into a live feed in your head, and nothing on the page corrects you. Short interest belongs with the other headline figures that are precise about something other than what they get read as, which is the territory revenue backlog covers from the accounting side.
Two markets, two clocks. The work is knowing which one you are reading.
What the number can carry, and what it cannot
So what is a short interest figure actually good for, once you have stopped asking it to be current?
| Short interest | |
|---|---|
| What it measures | Open short positions in customer and proprietary accounts at FINRA member firms, as of a designated settlement date |
| What it omits | Short exposure held through swaps or options delta, and any position outside FINRA's member perimeter |
| How old it is when you read it | Published on the seventh business day after the settlement date, which runs longer in calendar days across holidays and weekends |
Structural questions, it answers well. Whether a security carries a large reported short position against its own history is a question about a level. A level measured on a two-week cycle is still a level. Whether that figure has been building or draining across several prints is a question about direction, and direction survives a lag that a snapshot does not. Both readings live comfortably inside what the data is.
Point-in-time questions are where it breaks. Anything phrased as "how much of this is short right now" is asking a settlement-date photograph for a live quote. The number cannot know. Neither can you. That is not a reason to discard it, but a reason to size a position on what the figure supports rather than on what its precision implies.
Eunha's last point outlasts the rule, and it is not really about short interest. Every published number has a production process, and the process leaves fingerprints: a date it was struck, a boundary around what it could see, an institution that decided what was worth compelling. The figure on the screen is the output. The rule that produced it is what tells you the output's meaning, and it is almost always available to read. Short interest just happens to be a case where the regulator published both.
Reading the date before the size takes about thirty seconds, so go and do it on something. Take a number you would have sized a position on this week and locate the day it was struck before you look at its direction. Then open a position on one of the 35 tokenized stocks in the Kodex simulator and hold it through a print, tracking what you actually knew going in against what you knew an hour later. Being wrong about that gap costs you a line in a $5,000 paper account and nothing else.
