Why a stock's float differs from one screener to the next

How it works

Float starts as a subtraction. Take the shares outstanding, which the issuer reports, and remove the shares that are held for some reason other than sale: insiders, founders, a parent company, a government, a private-equity fund, anything under a lock-up. What is left is the float, and a float-based ratio such as percent of float short divides by it.

Only the first number is a reported fact. The second is a judgement about who is strategic, made by whoever publishes the float, using ownership filings that arrive at different times. No SEC rule, FINRA rule or exchange rule defines float, so there are as many floats as there are rulebooks, and three of them are published:

1. The issuer's own, on the cover of its 10-K.
2. An index provider's, such as S&P Dow Jones Indices or FTSE Russell, set out in a methodology.
3. A data vendor's, which may follow one of the above, mix them, or use its own list, and rarely says which.

A screener shows one of these. The one beside it shows another. Neither has made a mistake.

The share count underneath is its own problem, covered in why two screeners disagree: which classes are counted, and as of which date. This page is about what is taken away from it.

The issuer's version: non-affiliates on the 10-K cover

The closest thing to an official float is a sentence every annual report must carry. The Form 10-K cover asks the registrant to "state the aggregate market value of the voting and non-voting common equity held by non-affiliates", at the price on "the last business day of the registrant's most recently completed second fiscal quarter". If deciding who is an affiliate would take "unreasonable effort and expense", the issuer may use "assumptions reasonable under the circumstances", provided it states them.

The test of who is an affiliate is control, not size. Rule 144 defines an affiliate of an issuer as "a person that directly, or indirectly through one or more intermediaries, controls, or is controlled by, or is under common control with, such issuer". The rule sets no percentage, so the line is drawn company by company.

Tesla's 10-K for 2025, filed on 29 January 2026, shows what the sentence produces. Non-affiliate holdings were worth $892.93 billion at the close on 30 June 2025. "Shares of Common Stock held by each executive officer and director have been excluded in that such persons may be deemed to be affiliates", and the company adds that "this determination of affiliate status is not necessarily a conclusive determination for other purposes". The share count on the same cover, 3,752,431,984, is as of 23 January 2026.

Three things follow. The figure is a dollar value, so turning it into a share float means dividing by a price from a date nobody else is using. It is about seven months older than the share count printed beside it. And it is the issuer's own reading of who controls it, offered with a caveat.

Two index rulebooks, two answers

Index providers need a float to weight constituents, and they publish their rules. Comparing two of them shows how far apart reasonable rules land.

S&P Dow Jones Indices, in the methodology dated September 2026, computes an Investable Weight Factor, available float shares divided by total shares outstanding. Shareholders are "not judged individually based on their intentions"; they are grouped by type. The excluded types include officers and directors and related individuals whose holdings are disclosed, private equity and venture capital firms, asset managers and insurers with direct board representation, another publicly traded company, holders of restricted shares, company-sponsored employee plans, foundations or family trusts associated with the company, government entities other than pension funds, sovereign wealth funds, and "any individual person listed as a 5% or greater stakeholder". It also applies "a 5% minimum threshold for strategic blocks": in its own example, officers and directors holding 3% with no other strategic holder at 5% or more leave the factor at 1.00. Shares held through ADRs and similar receipts are normally in the float.

FTSE Russell, in its Free Float Restrictions ground rules, version 3.6 of October 2026, sets a threshold per shareholder type. Corporations, employee share plans, foundations, governments, treasury shares, and private equity and venture capital are restricted at any size. "All individual holders disclosed within corporate filings will be restricted, regardless of amount", officers and directors included. A single portfolio holding by a bank, hedge fund, adviser, pension fund, mutual fund or ETF is restricted only at 30% or more, and a sovereign wealth fund at 10% or more. Shares under lock-up are removed; nominee holdings are "typically regarded as free float".

Run one shareholder register through both and they diverge:

The two methodologies are written for index weighting, not for a screener column, and a vendor may follow neither. But they are the two most detailed public rulebooks for the same subtraction, and they disagree on ordinary holders.

The clock: when float is recomputed

A float is also a date, and the dates are set by the rulebook.

The 10-K figure moves once a year. S&P DJI reviews Investable Weight Factors annually, applied before the open on the Monday after the third Friday of September for most of its indices, with a cut-off for public information five weeks before the third Friday of the review month. At quarterly reviews in between, it changes a factor only when shares outstanding move by at least 5% and the factor itself by at least 5%. At the annual review every factor of 0.96 or above is rounded to 1.00, so a company that is 97% float by the rules is carried as 100%.

Rulebooks change too. On 12 August 2026 S&P DJI added regional rules for Brazil, Indonesia and Turkey, which it called a clarification, taking effect with the September 2026 reconstitution. A float series that spans a methodology change is two series.

Restricted stock creates the steps. Under Rule 144, shares bought from a reporting issuer in an unregistered transaction must be held at least six months before public resale, and an affiliate's sales in any three months are capped at the greater of 1% of the class outstanding and the average weekly trading volume over the four preceding calendar weeks. Lock-ups after an offering add contractual dates on top. When those dates pass, shares become saleable, and every provider moves them into float on its own schedule. FTSE Russell implements lock-up expiries at the next quarterly review and, where the released holders are not identified, adds them pro rata at the company's post-lock-up free float percentage: in its own example, 60 of 100 released shares. S&P DJI generally reviews lock-ups with its annual reconstitution unless a corporate event triggers an earlier review.

So for a recent IPO or a company coming out of a lock-up, the float on two screens can differ by the whole released block for weeks or months. That is the period in which percent of float short is most often quoted.

Percent of float short: two numbers, two dates

The numerator of percent of float short is usually short interest, the open short positions broker-dealers report to FINRA for two settlement dates a month. Where it comes from, when it is published and what it leaves out is in short interest, short volume, fails and borrow fees. The denominator is whichever float the site uses.

The arithmetic is unforgiving. Take a company with 100 million shares outstanding and 10 million shares short:

Same short position, three headlines. Add the date mismatch, a short interest figure for one settlement date divided by a float last recomputed months earlier, and two honest sites can differ by a factor of two. The numerator can differ too: ORTEX estimates short interest daily from securities-lending inventory alongside the official figure, and other products in insider and 13F trackers carry their own estimates.

Days to cover does not use float at all. It divides short interest by volume, which is why it can move in the opposite direction from percent of float short on the same day.

What you can do about it

Find out which denominator the column uses. Float or shares outstanding, and if float, whose. A vendor that names its method lets you compare; one that does not has given you a number you cannot reconcile with any other.

Reproduce the float from the 10-K once. Open the latest annual report through SEC EDGAR, divide the non-affiliate market value by the closing price on the date it names, and compare the result with the screener's float. If they are close, the screener is probably excluding insiders only. If the screener's float is much smaller, it is removing large holders the issuer does not treat as affiliates. The large holders are listed in the proxy statement; Tesla's 10-K, for one, incorporates it into Part III.

Check the date on both numbers. A percent of float short is only as current as the older of its two inputs. Write down the short interest settlement date and, if the site gives it, the float's as-of date. If the float is older than the last offering, buyback or lock-up expiry, the ratio describes a company that no longer exists.

Treat recent IPOs and post-lock-up stocks separately. In the months after an offering, float moves in steps as restrictions lapse, and each provider takes the step on a different day. Compute the ratio against shares outstanding as well; the truth for your purpose is somewhere between the two.

When an index weight matters, use the index provider's own rules. If the question is how much of a stock an S&P or FTSE Russell index fund must hold, only that provider's factor answers it, and the methodologies above say how it is built and when it changes.

Compare like with like across tools. Two screeners in charting platforms and screeners that disagree on percent of float short are usually both right about different things. Pick one source for the ratio, note its definition, and stick with it, so a change in the number is a change in the company rather than a change of rulebook.