Fails to Deliver

What a fail to deliver actually is

When you buy a share, the seller has two business days to deliver it. Usually they do. When they do not, the trade is a fail to deliver, and the National Securities Clearing Corporation carries the open balance until it is resolved. The SEC publishes those balances by security, twice a month, going back to 2009.

The reasons are mundane and mostly administrative:

• Corporate actions. A merger, a spin-off, a name change, a symbol change. Positions in flight across the record date break, and every one of them fails until the paperwork catches up. The very largest balances in this window are exactly this — hundred-million-share balances in megacaps, for a fortnight, then gone.
• Mechanical errors. A wrong account, a late instruction, a transfer agent that did not process a certificate in time.
• Market-maker exemptions. Registered market makers get extra time to deliver in the course of bona fide market making, so their hedging shows up here briefly and legitimately.
• Illiquid borrow. A security genuinely hard to locate, where a short seller could not source the stock. This is the case people mean, and it is a minority of the file.

What it is not

A fail balance is not a count of naked short sales, and the SEC says so directly on the page this data comes from. Under Regulation SHO, a fail of any size in any security triggers a mandatory close-out obligation; a security that sustains fails past a threshold goes on the Reg SHO restricted list, which carries harder requirements still. The regime is designed so that fails resolve, and the data shows that they do.

The other frequent error is arithmetic. These are balances, not flows. A security showing a hundred thousand shares on five consecutive reported dates has not failed five hundred thousand shares — it likely has one unresolved position that persisted. Adding the column up produces a number that means nothing, and that sum is the basis of most of the alarming figures that circulate.

Why this page cannot show a streak

The SEC states explicitly that the age of a fail cannot be determined from the file, and that today's balance is not necessarily composed of the same fails as the last report's. So “days of consecutive failing” — a headline metric on sites that publish this same file — is not derivable from it. What is derivable is how many reported dates carried a balance at all, which is the column above. A security at 73 of 73 has had an open balance at every reading; whether that is one long-lived break or many short ones, this data cannot say.

Reading it usefully

• Scale to the security. A million-share balance is noise in something trading forty million shares a day and extraordinary in something trading fifty thousand. The absolute number on its own says very little.
• Persistence beats size. A one-off spike is almost always a corporate action. A small balance present at nearly every reading, in a small security, is the more interesting pattern — it suggests the stock is genuinely difficult to source.
• Read it with the other two. Fails alongside heavy short interest and elevated short volume is a coherent picture. Fails alone, in a name with no short interest, is a settlement problem and nothing more.

Where this comes from

The SEC publishes the CNS fails-to-deliver file free and without restriction. Each archive here is checked against both trailer records the file carries — its record count and its total share quantity — and any archive that does not reconcile on both is skipped rather than partially loaded. A recent window is stored rather than the full archive back to 2009, which runs to tens of millions of rows.

Every security's own page carries its fails alongside its filings, ownership and short interest — see any ticker under security pages, or the rest of the Markets section.