The By-the-Glass Cliff: Where Restaurant Wine Markup Stops Buying Anything

Wine by the glass used to be the bottom of the list, and there was a physical reason for it rather than a snobbish one. Oxygen. From the moment a bottle is opened the wine starts to change, and a restaurant pouring the last glass three days after the first is selling two different products at one price. For most of the history of restaurant wine service that constraint decided the category. The glass pour was the house red: bought cheaply, sold quickly, not expected to survive scrutiny. A wine worth talking about was sold by the bottle, because the bottle was the only unit that could guarantee what was in it.
The category changed when the oxygen problem became a hardware problem. Cruvinet, which dates its own business to 1978, built cabinets that blanketed opened bottles under nitrogen and dispensed them through taps, and across the 1980s and into the 1990s that class of machine and its competitors made a wide by-the-glass list operationally possible for the first time. Before preservation hardware, a glass list was capped by what the room could reliably drink through in a night or two. After it, the cap moved, and the interesting question stopped being what a restaurant could keep open and became what it would charge for the privilege.
The pour itself got codified unevenly, which is worth knowing before you read any glass list. In England and Wales the Weights and Measures (Intoxicating Liquor) Order 2011 requires wine sold by the glass to be served in 125ml or 175ml measures, or multiples of them, and requires the house to display which measures it serves. That is why a British wine list carries numbers and an American one usually does not. In the United States there is no equivalent federal specified quantity. The five-ounce pour, roughly 150ml, is a convention of the trade rather than a rule, and a convention is a thing a house can quietly move by half an ounce without telling anyone.
The last piece of the modern picture arrived in 2013, when Greg Lambrecht, an engineer out of medical devices, put a hollow needle through a cork, pushed argon in behind the wine as it left, and let the cork reseal itself. The Coravin took the open-bottle clock off the top of a list. A restaurant can now pour one glass of something serious without committing the rest of the bottle to a two-day countdown, which is the reason glass lists at the high end look nothing like they did in 2005.
That is the craft history, and all of that hardware is real cost. It is the honest half of the markup, so it is worth taking seriously before taking anything apart.
The trade's own rule of thumb for by-the-glass pricing is blunt and widely published: charge per glass what the whole bottle cost you wholesale. A 750ml bottle yields about five five-ounce pours allowing for loss, so a house buying at twelve dollars and pouring at twelve dollars is returning something like four to five times its wholesale cost across that bottle, against a bottle-list markup that more commonly runs two and a half to three times. Stated plainly: the first glass sold pays for the bottle, and everything after it is margin against the risk that the rest of the bottle never sells at all.
That risk is not imaginary, and it is not the only cost. Preservation systems and the gas that feeds them. Glassware, which breaks, and good glassware, which breaks expensively. Pour discipline, which is the quiet one: a bartender with a heavy hand who turns five pours into four has raised the cost of goods on that bottle by twenty-five percent and nobody in the building will notice for a month. Staff who can describe what they are pouring. Storage that actually holds temperature. A by-the-glass program is genuinely more expensive per milliliter to run than a bottle program, and a list priced to reflect that is not gouging anybody.
The question is where reflecting the cost stops and the rest begins.
It stops sooner than most lists imply. Below about nine dollars a glass in an American dining room the arithmetic answers for you. At the trade's own pricing rule, a nine dollar glass is coming out of a bottle the house bought for around nine dollars, and a nine dollar wholesale bottle is commodity wine. It can be clean, sound and perfectly pleasant. It is not going to have a producer whose name is worth printing, and the menu usually reflects that by not printing one.
Between roughly nine and sixteen dollars the curve climbs hard. This is the band where the wholesale bottle cost crosses the line at which a real producer, a stated vintage and a specific place become available at all, and the difference between a nine dollar glass and a sixteen dollar glass is a difference you can taste without concentrating on it.
Around sixteen dollars the curve flattens, and that is the cliff. At that price the house is pouring from a bottle it bought for something like sixteen dollars wholesale, which in almost every category on a normal list buys a genuinely well-made wine from a named producer in a named place. Above that line the incremental dollar increasingly buys three things that are not wine: the preservation hardware that makes a rare pour possible at all, the prestige of a list that can offer it, and the insurance premium on a slow-moving bottle that might be poured out across a week.
There is one number that should temper all of that, and leaving it out would be dishonest. In the largest systematic study of restaurant wine pricing this piece could locate, David de Meza and Vikram Pathania examined 470 wine menus from 249 London restaurants covering 6,335 wines, and found that a 250ml glass carried less than seven percent more markup per unit of volume than the bottle did. That is a far smaller by-the-glass penalty than the American trade rule of thumb implies. Whether it holds outside London, and at five-ounce pours rather than 250ml ones, is not established, and this piece does not claim it does. What it does establish is that anybody asserting a flat, universal by-the-glass rip-off is asserting something that has been measured once and did not come out that way.
Which brings us to the folklore, since it comes from the same study.
The story is that restaurants load their heaviest markup onto the second-cheapest bottle on the list, because diners are too embarrassed to order the cheapest and will reliably take exactly one step up. It is repeated everywhere as a fact, almost always by people who have never seen it tested and are not aware that it ever was. De Meza and Pathania tested it, and published the result in Economics Letters in 2021. The markup on the second-cheapest wine came out significantly below the markup on the next four wines up the list. The peak markup sat near the median wine, in the middle of the list, which is where standard pricing theory for a product line would put it and where the psychological-exploitation story would not.
One study, one city, bottle prices rather than glass pours, and roughly two thirds of the wines matched to a retail comparison. It is not the last word on the subject. It is, as far as this piece could determine, the only word, and it says the opposite of the thing everyone repeats. Until somebody measures it in another market and gets a different answer, the second-cheapest-bottle rule is not a finding. It is a thing people say at tables. The useful version of the advice, on the only evidence there is, points at the middle of the list rather than the bottom of it.
The practical read on a glass list is not the price anyway. It is the specificity. A menu that names the producer and the vintage is telling you it is pouring from a bottle it would be willing to sell you whole. A menu that says Chardonnay, California, twenty-two dollars is telling you something as well, and what it is telling you is that the producer changes with whatever the distributor had. A printed pour size means the house has decided what a glass is and can be held to it. And a glass list short enough that the room can plausibly drink through it in two nights means the wine in your glass was opened recently, which affects what is in the glass considerably more than four dollars of markup does.
Where it came from
- 1978
Nitrogen makes a real glass list possible
Cruvinet dates its business to 1978, building cabinets that held opened bottles under nitrogen and dispensed them through taps. Across the 1980s and 1990s that class of hardware lifted the ceiling on a by-the-glass list, which until then was capped by what a room could drink through before oxidation did its work. The date here is the company's own account of itself rather than an independent record.
Source: Cruvinet Winebar Company, About (company self-reported founding date)
- 2011
Britain fixes the size of a glass, America does not
The Weights and Measures (Intoxicating Liquor) Order 2011 requires wine sold by the glass in England and Wales to be served in 125ml or 175ml measures or multiples of them, and requires the premises to display which it serves. There is no equivalent federal specified quantity in the United States, where the five-ounce pour is trade convention rather than law.
- 2013
Coravin removes the open-bottle clock
Greg Lambrecht, an engineer from medical devices, brought the Coravin to the US market in 2013: a hollow needle through the cork, argon pushed in behind the wine, and the cork resealing itself afterwards. It let a restaurant pour a single glass of something serious without committing the remainder of the bottle to a countdown, which is why high-end glass lists no longer resemble the ones from a decade earlier.
Source: Coravin, company account of the March 2013 US launch and the argon needle system
- 2021
The second-cheapest story finally gets tested
David de Meza and Vikram Pathania examined 470 wine menus from 249 London restaurants covering 6,335 wines. The markup on the second-cheapest wine came out significantly below the markup on the next four wines up the list, with the peak markup near the median wine. In the same dataset a 250ml glass carried under seven percent more markup per unit volume than the bottle. Bottle prices in one city, so not the last word, but it is the only systematic test of the folklore this piece could locate.
How it is made
A by-the-glass program is genuinely more expensive per millilitre to run than a bottle program: preservation hardware and gas, glassware that breaks, pour discipline, and the standing risk that the back half of an open bottle never sells. The markup up to the cliff is paying for that. Past the cliff it is paying for the list.
Restaurant wine by the glass
What to buy
Recommendations rest on verifiable economics, and we have not tasted anything: no blind tasting was run, no restaurant is named and no wine is recommended. The cliff derives from three things — the trade's own published by-the-glass pricing rule, in which the glass price approximates the wholesale cost of the whole bottle across roughly five pours; the pour conventions and the English and Welsh statutory measures; and the one systematic study of restaurant wine markups located for this piece, which examined bottle lists in a single city rather than glass pours across markets. Price bands are editorial judgement about what a given wholesale cost makes available, not a price survey.
A glass at or just above the cliff, from a producer and vintage named on the menu
At the trade's own pricing rule that glass comes from a bottle the house bought at roughly sixteen dollars wholesale, which in nearly every category buys a well-made wine from a named place. Above it the money goes to the list rather than the glass.
The third glass
If the glass price approximates the wholesale bottle cost, two or three glasses cost about what the same bottle costs on the same list. If the table will drink three, the bottle is usually the cheaper unit and the wine is fresher, because it was opened for you. This follows from the American trade rule of thumb, and the London finding recorded in the history block is a genuine caution against treating it as universal.
The second-cheapest-bottle rule
The only systematic test of it found the markup on the second-cheapest wine significantly below the next four up the list, with the peak markup near the median. One city, bottle prices, not the last word. But the folklore has never been supported by evidence anywhere, and it has now been contradicted once. If you want to dodge the heaviest markup on the evidence available, avoid the middle of the list, not the bottom of it.
- How this was made
- Recommendations rest on verifiable economics, and we have not tasted anything: no blind tasting was run, no restaurant is named and no wine is recommended. The cliff derives from three things — the trade's own published by-the-glass pricing rule, in which the glass price approximates the wholesale cost of the whole bottle across roughly five pours; the pour conventions and the English and Welsh statutory measures; and the one systematic study of restaurant wine markups located for this piece, which examined bottle lists in a single city rather than glass pours across markets. Price bands are editorial judgement about what a given wholesale cost makes available, not a price survey.
- Sources
- 4 consulted, 4 primary
- Price bands
- Editorial judgement, not a formal price survey.
- Fact-check
- Passed
- Standards
- Passed
- Edited & approved
- Josh Perry, August 5, 2026