Fat Finger Error

A fat finger error is a human input mistake in financial markets: an order or payment entered with the wrong price, the wrong quantity, the wrong instrument or the wrong side, buy for sell. The name suggests a finger pressing the wrong key or one key too many. The typical case is a figure that is too large by several zeros, or a price and a quantity typed into each other's fields.

Such errors are a form of operational risk. In electronic markets an order reaches the market the instant it is submitted, so a slip that a colleague or a floor broker might once have caught can be executed in full within seconds. Most errors are small and are absorbed quietly. A few have cost hundreds of millions of dollars and moved whole markets.

Common forms include entering an amount in units when thousands were meant, transposing digits, confusing the price and size fields, selecting the wrong security from a list, and submitting the same order twice. Errors also arise one step removed from the keyboard. A wrong parameter typed into an algorithm, a spreadsheet uploaded with a column out of place, or a test order released into the live market is the same kind of mistake with larger reach.

The consequences depend on the order type and the market. A large market order sweeps through the order book and trades at whatever prices are available, which in a thin market can be very far from the last price. A limit order entered at a badly wrong price trades at once against every better-priced order. The sudden move can trigger other traders' stop orders and automated strategies, turning one mistake into a brief flash crash.

• Mizuho Securities, 2005. A trader at the Japanese broker intended to sell one share of a newly listed company, J-Com, at 610,000 yen. The order was entered as 610,000 shares at 1 yen. The exchange's system did not accept the cancellation, and the firm lost about 40 billion yen.
• Deutsche Bank, 2015. The bank mistakenly paid about $6 billion to a hedge fund client when a gross figure was processed instead of a net one. The money was recovered the next day.
• Samsung Securities, 2018. The South Korean broker meant to pay employees a dividend of 1,000 won per share and instead issued 1,000 shares per share, creating billions of shares that did not exist. Some employees sold them before the error was reversed.
• Citigroup, 2020 and 2022. In 2020 the bank, acting as loan agent for Revlon, sent lenders about $900 million of its own money in error. In May 2022 a London trader preparing a basket of shares worth $58 million created one of $444 billion; part was blocked, but about $1.4 billion of shares were sold on European exchanges, causing a sharp fall in several indices. UK regulators fined the firm about £61.6 million in 2024.

Firms and venues defend against input errors in layers.

• Order entry limits. Trading systems reject or query orders above a maximum quantity or value, and orders priced too far from the current market. A soft limit shows a warning that the trader can override. A hard limit cannot be overridden.
• Pre-trade risk checks. A risk management system independent of the trader tests each order against position, credit and loss limits. These are central to algorithmic trading controls.
• Second approval. Large payments and unusual orders require sign-off by a second person, the maker-checker principle.
• Venue safeguards. Exchanges apply price bands, a volatility interruption or a circuit breaker that pauses trading when prices move too far too fast.
• Kill switches. A firm or venue can cancel all of a participant's open orders and block new ones at once.

Investigations into large incidents commonly find that controls existed but were set too loosely, that warnings could be dismissed in bulk, or that many pop-up alerts had taught staff to click through without reading. Limits need to be tight enough to bind and few enough to be taken seriously.

Once an erroneous order has traded, the trades are generally binding. Exchanges have rules for cancelling or repricing trades executed at prices far from the market, often called clearly erroneous or error trade policies, but these apply only within set thresholds and time limits and are used sparingly, since other participants traded in good faith. The firm at fault usually has to trade out of the unwanted position at a loss.

For mistaken payments the position depends on local law. Recipients are generally obliged to return money sent in error, though the Revlon case showed that recovery can be contested when the recipient was owed the amount.

Regulators treat input errors as a failure of systems and controls and not as bad luck. In the United States the Securities and Exchange Commission market access rule requires brokers to have controls that prevent erroneous orders, including those exceeding price or size limits. In the European Union, MiFID II sets similar requirements for firms and trading venues. In the United Kingdom the Financial Conduct Authority has fined firms for weak trading controls following such incidents. Clients given direct market access fall under the same controls, applied by the broker.

• Financial Conduct Authority
• U.S. Securities and Exchange Commission