Why an over-the-counter market has to manufacture a benchmark
Gold is not primarily an exchange traded asset. The bulk of trading happens bilaterally, between banks and their clients, at prices negotiated privately. There is no central limit order book for the main market and therefore no closing bell and no official settlement price falling out of the day naturally. An exchange produces a benchmark as a by-product of its own operation. An over-the-counter market has to build one deliberately.
The requirement is not academic. Supply contracts, structured products, exchange traded funds, mining agreements and official reserve valuations all need a reference that both sides accept in advance and neither side can influence. That reference has to be transparent in method, produced at a stated time, and based on real transactable interest rather than on indicative quotes. The auction exists to satisfy exactly that list of demands, and its design choices make much more sense when you read them as answers to a governance problem rather than as a trading mechanism.
How the auction actually runs
It is an electronic auction with a small group of accredited direct participants, each of which can submit orders on its own account and on behalf of clients. The process is iterative rather than a single sealed bid.
An opening price is published. Participants enter the quantity of gold they wish to buy or sell at that price within a short timed round. The system then computes the net imbalance, the difference between total buying and total selling interest. If buying exceeds selling by more than a permitted tolerance, the price is moved up for the next round. If selling dominates, it is moved down. Participants resubmit at the new price, the imbalance is recalculated, and the loop repeats.
The auction ends when the imbalance falls within tolerance, and that price becomes the published benchmark. Notice what this means: the printed price is the level at which buying and selling interest roughly balanced among that group at that moment. It is a clearing price, derived from the orders in the room, which is both its strength and the boundary of what it can tell you.
Who actually needs the printed price
Almost nobody trading intraday cares about the benchmark, and almost everybody dealing in physical metal or long dated contracts does. The users fall into recognisable groups.
- Producers and refiners, who sell output under contracts priced off a published reference rather than off a negotiated quote.
- Fabricators and jewellery manufacturers, buying metal on the same basis.
- Funds holding allocated bars, which must value their holdings each day using a defensible number.
- Issuers of structured products, whose payoffs are defined against a named benchmark so that neither party can argue about the price afterwards.
- Official institutions, marking reserves.
The common thread is that each of these users needs a price they did not choose. That is the entire point. A benchmark is valuable precisely because it is produced by a published process at a published time, so neither counterparty can claim the other picked a convenient moment. It is infrastructure for settling obligations, not a view on gold.
Why activity clusters around the auction windows
If a large volume of real business is going to be transacted at a price determined in a few minutes, participants with exposure to that price will manage it around those minutes. A dealer who has agreed to fill a client at the benchmark has a position whose value is fixed by the auction outcome, and they will hedge accordingly before, during and after.
The result is a reliable concentration of genuine two way activity at two points in the London day, one in the late morning and one in the mid afternoon. The afternoon window carries more weight because it overlaps with United States market hours and is the reference most widely written into contracts. These are not mysterious events. They are the moments when a lot of real metal business is priced, which is exactly why the surrounding minutes see heavier flow than the hours either side of them. Anyone mapping out when gold is liveliest will find this sits naturally alongside the broader rhythm covered in the London session on gold.
What it looks like on a chart, told honestly
Sometimes you see a short burst of volatility, a push in one direction and a partial retrace once the print is done. Sometimes you see a tight, orderly few minutes and nothing worth noting. Both outcomes are common, and that is the problem with the way this event is usually taught.
There is no dependable pattern to trade here. The auction does not reveal direction, because its output is a balancing price rather than an expression of appetite, and the hedging around it flows in whichever direction the day's business happens to require. Strategies built on an assumed spike and reversal rely on the hedging being consistently one sided, and there is no mechanism that would make it so. Some days the accumulated client orders are to buy, some days to sell.
What you can rely on is the character of the window rather than its direction: more flow, faster quotes, and a short period where price can travel further than the surrounding context suggests. That is a reason to be careful, not a reason to position.
Month end and quarter end raise the stakes
The importance of a particular print varies enormously depending on the date. On an ordinary Tuesday it settles the day's physical business. On the last business day of a month or quarter, the same number is used to value portfolios, mark fund holdings, strike product payoffs and set reported performance for a large number of institutions at once.
More value therefore depends on one print, and more participants have a genuine interest in the price at which it lands. This does not imply anything improper. It means the volume of real business concentrated into that window is larger, so the flow around it is heavier and the market can be less forgiving. The same calendar effects that drive the broader patterns in month end flows in gold reach their sharpest point at the valuation print, which is a sensible reason to avoid carrying marginal positions into those particular windows.
What it means for your fills
A retail trader cannot participate in the auction, cannot trade at the benchmark, and will not see the imbalance. Your orders are filled at a dealer quote with a spread on top, and during the auction windows that spread can widen as the dealer manages its own exposure to a price it does not yet know.
The practical consequences are mundane and worth respecting. Market orders placed into those minutes can fill further from the screen price than usual. Stops sitting just beyond obvious levels are more likely to be reached by a brief, flow driven extension and then left behind. Limit orders fare better than market orders. None of this is unique to the auction, it is the same arithmetic of execution cost described in the reality of spread costs, simply concentrated into a predictable part of the day.
What the benchmark cannot tell you
It is a settlement reference, so it carries no forecast. It describes where buying and selling balanced among a defined group during a defined few minutes, which is a statement about that moment and nothing else. A print above or below the prevailing market price reflects the orders submitted, not an institutional opinion about the next move.
Two further limits are worth stating plainly. First, the auction covers a slice of the market, substantial but not the whole of it, so it is not a census of global gold demand. Second, the published number is the outcome of a governed process, not a discovery about value, and treating it as a level that price must respect afterwards is reading significance into bookkeeping. Mark the windows on your live chart so you know when flow is likely to thicken, then let structure and your own rules decide everything else. Knowing when the market gets busy is useful. Believing the busy moment knows something is not.
FAQ
How often is the London gold benchmark set?
Twice on each business day, once in the late London morning and once in the mid afternoon. The afternoon print overlaps with United States hours and is the one most commonly written into contracts and used for valuations, which is why more real business depends on it than on the morning one.
Is the auction the same thing as the gold price?
No. It is one published reference price produced by a specific process at a specific time. Gold trades continuously around the clock at prices set bilaterally between dealers and clients, and the benchmark is a snapshot designed for settling contracts rather than a definitive market price.
Can I trade at the benchmark price?
Not as a retail trader. Only accredited direct participants submit orders into the auction, acting for themselves and for clients. Retail accounts fill at a dealer quote with a spread, and during the auction windows that spread can widen because the dealer is managing exposure to a price not yet determined.
Does price always spike during the auction?
No, and expecting it to is how traders lose money around these windows. Some auctions pass with barely a ripple while others produce a sharp move and partial retrace. The direction depends on which way the accumulated client business runs that day, and there is no mechanism making it consistently one sided.
Why does the month end print matter more?
Because the same number is used to value portfolios, mark fund holdings and strike product payoffs for a large number of institutions at once. More real business depends on that single print, so flow concentrated into the window is heavier and execution conditions around it can be less forgiving than usual.
ⓘ See these ideas on real price: open the free XAUUSD live chart.