Settlement date arrives and the securities you owe are not in the account. Talk me through what happens next.

Both legs or neither

Settlement is delivery versus payment, which exists so that neither party is ever exposed to having performed while the other has not. The consequence on a day when you cannot deliver is that nothing happens at all: the securities do not move, the cash does not move, and the buyer's money stays where it was. Your counterparty has not lost their trade and you have not been released from it.

That is the first thing to establish, because the instinct of most systems is to treat a settlement instruction that did not settle as a failed operation to be retried or, worse, rolled back. There is nothing to roll back. The trade remains a live obligation, the depository will present the instruction again on the next cycle, and it will keep doing so until it settles, is cancelled by both sides, or is closed out under the market's rules. Your record of it has to be able to say "intended to settle on this date, still open, failing for this reason, for this many days" as a normal state rather than as an error.

Find out why, because the remedies do not overlap

The reasons divide into three families, and confusing them wastes the only asset you have, which is hours.

The first is that you never had the stock. You sold something you expected to receive from a purchase of your own, and that purchase failed. This is a settlement chain, and it is the most common cause of fails in size: one failure at the top propagates down through everybody who was relying on the delivery. You cannot fix your own leg by fixing your bookkeeping, because the shortfall is real.

The second is that you have the stock but it is not available. It is out on loan and has not been recalled, it is pledged as collateral, or it is sitting in a different account, at a different custodian, or in a different depository from the one the instruction points at. Nothing is economically wrong; the position exists and is simply in the wrong place at the wrong moment. This family is entirely preventable and is the one an inventory projection is built to catch.

The third is that the instruction itself is wrong or unmatched. The counterparty's details, the place of settlement, or the standing settlement instruction do not agree, so the two sides never matched and the instruction was never going to settle regardless of inventory. This is a static-data failure wearing a settlement failure's clothes.

The control is the day before, not the day itself

By settlement morning your options are all expensive. The work that prevents fails happens earlier, and it consists of two things.

Pre-matching is the practice of confirming with the counterparty or through the depository, ahead of settlement date, that both instructions exist and agree. It converts the third family of causes into a problem you have a full day to fix, and it is why the settlement-cycle compression conversation is really a conversation about how many hours are left for pre-matching after affirmation.

The other is a forward projection of what you will owe and what you will hold. Start from settled positions, apply everything due to settle and everything you have lent, pledged or borrowed, and produce a per-line, per-place shortfall for each of the coming days. That projection is what tells the securities lending desk to recall a loan while a recall still has time to complete, and its accuracy is limited by the same reference data as everything else: the same security held under two identifiers in two places projects as two positions and hides the shortfall.

flowchart TD
    A[Intended settlement date] --> B{Securities available at place of settlement}
    B -->|Yes| C[Settles against payment]
    B -->|No| D[Fail recorded and instruction re-presented]
    D --> E[Deliver partially where permitted]
    D --> F[Borrow to cover]
    D --> G[Bought in by the receiving party]
    D --> H[Market claim for entitlements in the fail period]

The three remedies are not alternatives of equal standing. They are in ascending order of cost to you, and the last one is chosen by your counterparty rather than by you.

Choosing between the remedies

Partial settlement delivers what you actually hold and leaves the remainder open. It is the cheapest thing available and it is frequently switched off, either because the counterparty does not permit it or because somebody disabled it years ago to keep reconciliation tidy. Turning it on is one of the few unambiguous improvements available in this area, because a fail of the whole line when you held ninety per cent of it is a self-inflicted exposure.

Borrowing to cover means going to the securities lending market, paying a fee, delivering the borrowed stock, and returning it when your own inbound delivery arrives. The economics are simple and the decision is a comparison: the borrow fee against the cost of continuing to fail. That cost is not zero even ignoring penalties, because you are holding an unfunded position and your counterparty is not paying you.

Buy-in is what happens when the receiving party loses patience or is required to act. They go to the market, buy the securities they were owed, and charge you the difference between what they paid and the original price. The obligation is then closed at their execution rather than yours, and you carry whatever the market did in the interim. Where the trade is cleared, the clearing house's rulebook governs the timing and mechanics rather than the two counterparties, which is one of the practical differences between a cleared and an uncleared fail.

Several markets also operate daily cash penalties for late settlement, calculated on the value of the failing instruction and paid to the party that was kept waiting. The rate is set by the market's rules and I would not quote a figure without looking it up, but the design implication is the important part: a fail has a per-day price, so the age of a fail is a financial quantity and the ageing report is a cost report.

The economics do not stop at the two legs

A fail that spans a corporate action creates a second obligation. If the stock goes ex-dividend while your delivery is outstanding, the dividend is paid to whoever is on the register, which is not the buyer who should have owned it. A market claim moves the entitlement to the buyer, and depending on the market it is raised automatically by the depository or by the two parties bilaterally. This is why a fail cannot be treated as purely a timing matter: the buyer is entitled to the economics of ownership from the intended settlement date, and every entitlement that arises in the gap has to be found and passed on.

The answer to the sibling puzzle sits here too. Marking to market keeps the exposure controlled, but it does not make the buyer whole for what they were supposed to own, it does not release the funding they set aside, and it does nothing about the chain of parties downstream of them who are now failing for the same reason. Penalties and buy-ins exist to price a delay whose real cost falls on people who are not party to your trade.

A failed settlement is not a failed operation. It is an obligation that has not yet been discharged, priced by the day, propagating to everybody downstream of you, and the only cheap moment to deal with it was before it happened.