How Menu Scoring Works

A restaurant flips the question. At home you ask “what is this wine?”. At the table, surrounded by an unfamiliar list and a waiter hovering, you ask “which of these should I order?”. Menu scoring is built for that exact moment: it reads the whole list and does the comparing for you, so the homework is done before anyone comes back to take the order.

A different question

Scanning a single bottle identifies one wine. A wine list is a dozen decisions at once, often with names you do not know and prices that range from fair to optimistic. Menu scoring exists because picking from a list is a different job from knowing a label, and it deserves its own tool.

How it works

1. You photograph the wine list, the whole page rather than one entry.
2. Every wine on the list is read and scored, weighing what each bottle is, how it tends to show, and whether the price is reasonable for it.
3. The smart picks rise to the top, each with alternatives, so you can order with quiet confidence instead of defaulting to the second-cheapest red.

Where you tell it what you are eating, it factors the dish in, the same reasoning as cellar pairing. I put this to a real test, ordering by the app’s pick across a dinner, in can AI pick the perfect wine for dinner? on the blog.

Versus cellar pairing

The two food surfaces answer the same instinct in opposite contexts:

• Cellar pairing recommends from bottles you own, at home.
• Menu scoring recommends from bottles you do not own, at a restaurant.

One reads your shelves; the other reads the page in front of you. Together they cover “what should I drink?” wherever you happen to be.

Where it stops

• It is Premium, and iPhone and iPad today.
• Reading a list is harder than reading a label. Glare, cramped columns, handwriting, and unusual layouts all make extraction tougher than a clean front label.
• Foreign and stylised menus are the edge cases, exactly where the app is most likely to leave an entry out rather than guess it.

When a wine cannot be read or grounded, it is omitted, not invented. The wider list of what Sommo can and cannot know is on the limitations page.

In the app

Menu scoring lives on its own feature page, with screenshots of scored entries and their alternatives. It is built for people who eat out and want to order well, so see Sommo for restaurant diners. If you have only ever known Vivino’s crowd scores, this is the difference between a popularity vote and a read of the actual list.

Frequently asked.