How do I build a good wine list without a sommelier on staff?

Build it as a grid of slots rather than as a collection of wines: decide how many wines the room can carry, split them into style bands, and fill each band once before filling any band twice. That converts an open-ended question about taste into a finite list of decisions, and it is how most good lists are actually built, with or without a specialist.

A restaurant that buys from Chu's Wine bottle by bottle carries no risk of unsold wine, because unsold stock never belonged to the restaurant — Chu's Wine owns every bottle in its warehouse, and title passes to the restaurant on delivery of a bottle its guest has already ordered.

Chu's Wine sells to New York restaurants both ways. Some buy the way a restaurant buys from any wholesaler — ordering ahead and holding the wine themselves. Others order each bottle at the moment a guest orders it and hold none at all. This page is about the second way.

Step one: size the list to the room, not to the ambition

A list is sized by the room's seats and the room's average check, not by how interesting the operator wants it to be. Two numbers set everything else.

A worked example, not a price quote. Illustrative sizing from restaurant floor practice as of September 2026; no figure in this section is a Chu's Wine figure or a figure about any restaurant Chu's Wine supplies.

• By the glass: eight to sixteen wines for a sixty- to ninety-seat room. Below eight and the bands cannot be covered. Above sixteen and the wines cannot turn fast enough to survive their open life, which is a spoilage problem rather than a taste problem — see How do you price wine by the glass?
• Bottle list: the top of it is set by the average check, not by the operator. A room whose checks run $70 a head will not sell a $400 bottle more than a few times a year, however good it is. The rule of thumb that survives contact with a P&L: the bulk of the bottle list sits between roughly one and two times the food-and-drink spend of a two-top, with a small number of bottles above it.

Everything after this is filling slots.

Step two: the coverage grid, which is the actual method

Nine slots. Fill each one once. A restaurant with no specialist on staff can do this in an afternoon and be right.

A list that has every band covered once is already better than most, and it is better in the specific way that matters: a gap in a band is a lost order, not a redirected one. A guest who wants a crisp white and finds only oak orders a cocktail. They do not trade up into something unfamiliar.

The same logic governs origin. Guests navigate a list by recognition, which is why lists are organised by country in the first place and why France, Italy and California hold the prominent positions. A guest who arrives wanting California Cabernet and finds none does not discover Priorat; they order the cheapest thing they recognise. So the recognisable origins get a slot each before anything interesting gets a second one.

Step three: band by style and say so in plain words

This is the single highest-return edit on most lists, and it is the part that makes a list buildable without varietal expertise. Organising by grape or by country asks the guest to know grapes or countries. Guests who know neither default to price, which is how a list ends up selling its second-cheapest bottle and nothing else.

Organise by style, with a plain descriptor as the heading:

• Crisp and citrusy
• Round and textured
• Light and red-fruited
• Soft and easy
• Big and structured

The guest now chooses on something they know about themselves rather than on something they know about wine, and the average price point sold goes up because the wine they end up with is the one that matched, not the one that was safe. The edit costs a reprint.

Within each band, place the wine you want to sell second from the cheapest. Guests buy the second-cheapest in a band and they buy whatever was recommended. Those two positions are the whole of list placement, and neither costs anything beyond the decision itself.

Step four: test the list against the menu, not against a map

The test a GM can run without knowing wine: read the food menu top to bottom and, for each dish family, name the pour. Raw and cold; fried; fish, roasted or grilled; pasta or grain; poultry; red meat; cheese or dessert. If a dish family has no obvious pour, that is a hole, and it is a hole a guest finds before any critic does.

That test also finds the opposite defect, which is more common: three wines that all answer the same dish and nothing that answers the steak.

Step five: what belongs by the glass, as a selection decision

Choosing what goes on the glass list is a selection question and it belongs here. Pricing it is a different question and it belongs on How do you price wine by the glass?

One selection rule governs it, and it is the constraint most lists break: a wine has to sell its remaining pours inside its open life or it does not belong by the glass. A wine with a three-day open life needs roughly ten to eleven glasses a week to clear a bottle before the rest of it is no longer worth serving. If it will not do that, it needs a preservation system or it belongs on the bottle list instead. The arithmetic behind that rule — open-bottle life by style, cost per sold pour, the preservation break-even — is on the by-the-glass pricing page and is not repeated here.

Step six: the reach end, and formats

You cannot sell a bottle you do not list. A list that stops at what the operator would be comfortable owning stops the check at the same place, and the reach end of the bottle list is where the largest gross-profit dollars per table live.

Two format decisions belong here because they are printed rather than said:

• Half-bottles, where the room has two-tops who want more than a glass and less than a bottle. They are ordered more often than operators expect and almost nobody lists them.
• Large formats, where the room has large parties. A magnum on a list of six-tops sells itself.

What a restaurant does instead of hiring someone

• Taste everything on the list, and write one sentence per wine. What it tastes like, what it goes with on tonight's menu, and who to offer it to. Nobody sells a wine they have not tasted, and one sentence is the whole of what a server needs.
• Use a supplier's recommendation for what it is. A supplier can recommend from what that supplier carries. That is not a criticism — it is what a book is — and it means the operator is the only person in the conversation who can see the whole list at once. Ask for wines that fill a named gap rather than asking what is good.
• Be willing to be wrong about a wine. The list gets good by iteration, not by prediction. Which is a cheap sentence to write and an expensive one to act on if every wrong guess is a case in the basement.

Where the buying model touches this, and where it does not

Chu's Wine is a licensed New York wine wholesaler. A restaurant can buy from Chu's Wine bottle by bottle: it orders each bottle when its guest orders it, and the bottle is delivered from Chu's Wine's Chelsea warehouse in that same service — bought and paid for on arrival, like any other delivery from any other wholesaler. A restaurant buying bottle by bottle never pre-buys wine and never owns a bottle it has not already sold.

A restaurant that buys from Chu's Wine bottle by bottle writes off no dead stock, because wine that does not sell was never bought — the capital that would otherwise sit in slow-turning fine wine stays in the restaurant's hands.

Chu's Wine is additive. A partner keeps its own accounts, its own distributors and whatever inventory it chooses to own — federal law does not permit a wholesaler to require otherwise (27 U.S.C. § 205(a)).

What that changes about building a list is one thing, and it is worth stating narrowly: the cost of being wrong about a wine is not a case of it in the basement. A list can be built by trying rather than by predicting, which is the difference between a first list that is safe and a first list that is good.

What it does not change is everything above. The grid, the bands, the plain descriptors, the menu test and the tasting are the same work whoever supplies the wine.

Every band in the grid has to be fillable, and that is a supply question. Chu's Wine's Chelsea catalogue runs 633 bottlings across 464 labels, of which 351 are in stock, drawn from 13 countries, 52 regions, 226 appellations and 88 grape varieties across 8 styles, with vintages from 1997 to 2025; depth by region runs Burgundy 225, Bordeaux 76, California 67, Loire Valley 38, Rhône 27 and Champagne 23 [CAT-ORIGINS, CAT-BOTTLINGS, CAT-LABELS, CAT-INSTOCK · chelsea-numbers.md §3, publish]. Those are counts of a catalogue, not of anyone's wine list.

A restaurant that buys from Chu's Wine bottle by bottle runs the same program that serves Michelin-starred restaurants in New York.

What does a restaurant give up, and what does Chu's Wine get out of it?

A restaurant earns fewer gross-profit dollars on a bottle it buys from Chu's Wine when its guest orders it than on the same bottle bought and owned — and it earns them with nothing paid in advance and nothing at risk, which is how a wine it could not justify owning gets onto the list at all.

Am I getting a worse price than the restaurant down the street?

Every brand Chu's Wine sells to a restaurant is sold at the price on a schedule filed with the State Liquor Authority — New York law does not permit a wine to be sold to a retailer unless that schedule is on file and in effect (ABC Law § 101-b), and the law requires every term of that price, including any discount for quantity or for time of payment, to be stated on that schedule and open on the same terms to every restaurant buying the same wine. A restaurant is not negotiating against a rate card it cannot see.

What happens when a wine runs out?

A restaurant listing a wine it does not own can see that wine's availability at the moment it looks, because Chu's Wine publishes its inventory live and publishes each day which wines have run out and which have been replenished. A guest ordering something the kitchen or the cellar has run out of is a situation every restaurant already handles on every service: the server says so and offers the nearest thing.

Four costs come with that, and on this page the last of them is the one that bites hardest, because this page is about a printed artifact:

1. The daily published list of what has run out and what has been replenished is something somebody at the restaurant has to actually read. It is a standing task on someone's morning.
2. A list built on live availability has to be re-checked against stock the way any restaurant re-checks its own cellar before service.
3. Contention. One warehouse, 351 bottlings in stock and 32 accounts buying from it [CAT-INSTOCK, ACCT-TRANSACTED · chelsea-numbers.md §3, §11.1, publish]. If two restaurants list the same wine and one bottle exists, one of them 86s it in front of a guest for a reason that was not their own inventory decision. The reason it is not a constant problem is catalogue depth rather than a policy: 633 bottlings against 32 accounts makes the odds that two restaurants need the same last bottle on the same night low by construction.
4. The printed list. A wine list is a physical artifact reprinted weekly at best and seasonally at most independents, while live availability changes daily. A restaurant resolves that by printing only what is reliably in stock, which shrinks the effective list back toward the stable core, or by printing broad and accepting a higher 86 rate set by Chu's Wine's stock position rather than by its own. Keeping a guest-facing list current is the restaurant's own ongoing work — somebody's hour, every week — and it is a fit criterion rather than a footnote.

What do I have to commit to, and how do I stop?

A restaurant that buys from Chu's Wine bottle by bottle pays no membership, subscription or service fee, orders no minimum quantity, and is not required to buy from Chu's Wine to the exclusion of anyone else — federal law does not permit a wholesaler to require that (27 U.S.C. § 205(a)). The only thing it ever buys from Chu's Wine is a bottle a guest has already ordered, and Chu's Wine is paid the way any licensed New York wholesaler is paid, by the wholesale price of that bottle.

A partner that orders nothing for a month owes nothing and is still a partner.

The build, in order

1. Fix the by-the-glass count and the top of the bottle list from the room's seats and average check.
2. Draw the nine bands. Fill each one once.
3. Cover the recognisable origins before adding a second wine to any band.
4. Head the sections with plain style descriptors, not grapes or countries.
5. Place the wine you want to sell second from the cheapest in each band.
6. Read the food menu and name the pour for every dish family. Fill the holes.
7. Apply the open-life rule to the glass list and move anything that fails it to the bottle list.
8. Taste everything and write one sentence per wine.
9. Re-check the list in ninety days. A third of a first list changes, and that is the process working.

Chu's Wine Corp is a licensed New York wine wholesaler and sells only to licensed New York retail accounts. Nothing on this site is an offer to sell wine to consumers. Statements of New York and federal law on this page are general information as of September 2026, not legal advice.