Methodology
Principles
We do not score wines and do not rank wineries by how good they are. We collect verifiable facts about the producer, the price and the awards, and derive a category from them by published rules. The rules are the same for everyone: two people with this methodology and the same sources arrive at the same result.
The category is computed by a deterministic function, not by a language model. Feature weights and thresholds live in a versioned config in the repository, and the version is stored next to the producer's category — it is always visible which edition of the methodology produced the result. Weights cannot be changed quietly: a set of reference wineries with known categories is checked on every build, and a change that breaks it fails the build.
Silence in the sources is not evidence. Where data is missing, the producer never gets a lower category by default: the absence of confirmation says something about us and our sources, not about the company. But silence about one thing is not silence about everything — a winery whose vineyard area is not published is not thereby a winery we know nothing about, and we say what we do know.
Every conclusion opens up to the facts it rests on: each fact carries a value, a source, a confidence level and the date we recorded it. That is both a product feature and our legal position — we publish what can be observed, not guesses about what is hidden.
There is no single «wine score» here and there will not be one. Provenance, recognition and the position of the price are shown separately, each with its own explanation, and we give no tasting scores of our own: a single taster would be exactly the matter of taste this project is built away from.
Producer categories
A category answers two independent questions at once, and we compute them separately. The first is where the grapes come from: does the winery make wine from its own fruit, does it buy some in, or did we find no confirmation of a vineyard base at all. The second is scale. Keeping them apart matters because the first question is almost always answerable and the second often is not: vineyard area is published by a minority of wineries, and a single scale used to fall silent about a producer entirely just because that one number was missing.
The category is the projection of that pair. Own grapes at a small scale is a small family winery, at a larger one an estate. Own grapes plus bought-in fruit is a large producer with own vineyards; a profile of mass purchasing is an industrial producer. Where we found confirmation of neither vineyards nor full-cycle production, the category is a trading brand — defined through us and not through the company: these are producers for whom we were unable to find that confirmation. And where own winemaking is established but the scale is not, the category says exactly that: a winery with its own production. A category describes the production model, not the quality of the wine; we declare no public ladder of better and worse between them.
One category is worth reading carefully, because it is easy to confuse with a trading brand: winemaking from purchased grapes. It is used when the winery makes the wine itself but buys the fruit, and both halves are established — usually because the winery says so plainly («we don't have any productive vineyards yet»). A trading brand is a statement about our data: we found no confirmation. This one is a statement about a known fact, and no judgement is attached to it — for a young natural winemaker in Georgia, starting without land of your own and working with the fruit of small growers is an ordinary way to begin. A winery whose vineyard area we simply could not find never lands here: for that, the absence has to be confirmed, not merely unfound.
Every known feature adds a weighted contribution to one of two sums: for own winemaking, or for buying in grapes and bulk wine. Own vineyards from 100 ha weigh 3, above 10 ha weigh 2, a small holding of up to 10 ha weighs 1; vineyards mentioned without an area weigh 1; a production address in a wine region weighs 1 for own winemaking, a Tbilisi-only address weighs 1 for sourcing. Eight or more appellations in the range weigh 3, Kakheti and Racha-Lechkhumi in one price list weigh 2, spirits in the lineup weigh 1.
The origin axis is then decided by thresholds, not by a total score. A vineyard base counts as established by an area figure, by vineyards mentioned without one, or by full-cycle production in a wine region whose range shows no sign of mass purchasing. On top of that base the sourcing sum decides: from 5 it is mostly bought-in fruit, above 1 a mix, otherwise own grapes. Scale follows the area: up to 10 ha is small, up to 100 ha medium, from 100 ha large. Where the area is not published, a catalogue confined to a single region with almost no appellations, together with estate bottling or a named winemaker, still reads as a small winery — and is marked as an inference.
That mark is part of the answer, not a footnote to it. A category resting on a document — an area figure, a confirmed absence — is shown as documented; one resting on indirect evidence is shown with a dashed badge and says so in words. We would rather state a conclusion and its strength than stay silent and let the silence read as ignorance.
«Not enough data» remains, but now it means what it says: we could not establish where the grapes come from at all. It is not triggered by a missing hectare figure, and it is never a downgrade — silence in the sources still is not evidence.
A gap between the company's registration year and a «since 18XX» on the label does not affect the category at all — it is published as a separate fact. And the category itself never changes silently: a recomputation goes into the review queue for a person to confirm, and the decision stays in the classification history for good.
Facts and sources
A significant fact is not stored as a field on a card but as a statement: a value, a source, the date it was observed and a confidence level — state registry, producer statement, inference from catalogue data, or manual verification. On the producer's page that list opens under «What this is based on», with the source and the date next to each fact.
Classification uses features from outside sources — production address, own vineyard area, full-cycle production, the estate-bottled mark, a named winemaker — together with features counted from our own catalogue: the number of appellations and regions in the range, Kakheti and Racha side by side in one price list, spirits in the lineup.
Sources are ranked: registry first, then the label, the producer's own site, the retailer, everything else last. An individual source can be raised or lowered relative to its type — a fact checked by a person, for instance, should not lose to a shop scraper.
When sources disagree we do not quietly pick the right one. Both versions stay in the database with their sources and dates, the higher-ranked one wins, and the discrepancy is flagged for review — and, when it is worth the reader's attention, shown openly. «The label says since 1920, the company was registered in 2019» is content, not dirt in the data.
Re-importing the same value from the same source changes nothing; a new value from that source supersedes its own previous one, while statements from other sources are left untouched — their disagreement is settled by rank, not by overwriting.
From shop catalogues we take facts only: the name of the listing, the price, availability and the link. Retailers' own descriptions we do not copy.
Weighing awards
A medal by itself says little: competitions differ both in their rules and in the share of samples they award. We do not add medals up and do not derive a rating from them. Instead every competition is assigned a weight class, and that class is shown next to the medal everywhere the medal appears.
The class is assigned from four public facts: whether tasting is blind and the regulations are published; what share of entries receives a medal (if almost everyone is awarded, the medal stops telling wines apart); how the jury is composed and how independent it is from the entrants; and what entry costs, including whether marketing packages are sold together with the medal.
The basis for each class is published on the competition page, in plain words: we state what exactly we relied on. A wine with no medals is not considered worse than a wine with a medal from a pay-to-play competition.
A medal always belongs to a batch, not to the wine in general, so we show the vintage year of the awarded batch next to the year of the award. A medal from one year does not transfer to other vintages.
Sources spell the same metal differently — «Gold», «gold», «Gold Medal». On import the spelling is normalised to a single form; the metal itself is never changed, and an unfamiliar award keeps its own name rather than being forced into gold/silver/bronze.
Medal images are the competitions' own official marks. They are used nominatively — to point at the award itself — and always stand next to the competition name and the year; where we have no image, a neutral placeholder is drawn instead.
Found a medal listed incorrectly, or a competition weighed unfairly? Producers can send a correction and use the right of reply — see the section below.
- Major competition
A major international competition: blind tasting, published methodology, limited share of medals. Such a medal carries weight.
- Moderate weight
A visible competition with a partly published methodology. The medal counts, but is not equal to a major one.
- Pay-to-play
Paid entry and a high share of awards: such a medal says almost nothing about the wine.
- Unknown weight
Not enough public data about the competition — no weight assigned.
Price verdict
The verdict answers one question — is this wine expensive relative to comparable ones — and says nothing about its quality. It is computed from price observations: for each shop we take the latest recorded price among listings that were in stock, and the median of those is the wine's price.
That median is compared with the median of a cohort. The cohort is the wines of the same producer category and the same style class, and the style class is currently colour plus sweetness. If either is unknown, the style class is undefined and no verdict follows: «no data» is never treated as a value.
A deviation within ±15% of the cohort median is around the average, below it is below average, above it is above average. The tolerance and the other thresholds are configuration rather than a constant written into this text; on the wine page an expandable block shows the actual numbers — the wine's median, the cohort's median and the size of the cohort.
If the cohort holds fewer than eight wines, no verdict is issued at all: the card shows «Not enough data for a verdict» along with the actual cohort size. We would rather stay silent than draw a conclusion from three observations.
The verdict is also withheld when the wine falls outside its cohort's own price range — beyond the Tukey fences, one and a half interquartile ranges out from its quartiles. A reserve at seventeen times the cohort median is not the same wine sold dearer; it is a price tier the cohort does not contain, and calling it expensive would say nothing about a markup. Rare varieties and limited bottlings land here too. The range is computed from the cohort itself rather than as a fixed multiple of its median: a cohort with a wide spread tolerates a price a tight one does not.
Prices are collected from public shop pages and they age: every price on a card carries the date it was recorded. The price run happens once a week, so between our collection and your purchase the shelf price may well have moved.
Appellation authenticity risk
The risk flag belongs to the appellation, not to a company. It says that the name on the label guarantees nothing by itself, because the supply of that zone's wine on the market bears no relation to what the zone can physically produce. There are three states: no known authenticity issues, elevated risk, high risk.
Khvanchkara and Kindzmarauli work exactly that way: small zones with a chronic shortage of their own grape varieties and a volume on the shelves that those zones cannot yield. We do not claim that a particular bottle is fake — we say that in this zone the question is reasonable, and we list what to check.
What is worth checking is the producer's link to the zone: whether it owns vineyards there, whether it is registered in that area, whether the bottle names a specific vintage. A suspiciously low price for such a zone is a signal in itself — there is no explicable cheap Khvanchkara.
Next to the flag we show one more fact from the registry: whether the name is registered as a protected designation and since when. Some familiar names are not appellations at all but historical names of wines — such a word on a label is protected by nothing.
The flag is visible in three places: in the list of appellations, on the zone's own page with the explanation, and on the wine card next to the appellation, where the checklist opens. About companies we publish observable facts only — words like «counterfeit» and «fake» are never used towards a named producer.
Right of reply
The classification is computed from public data by published rules — which means it can be wrong precisely where public data is thin. A producer is entitled to dispute both a single fact and the category as a whole.
What convinces is a document, not a correspondence. The category is recomputed from facts, so anything that confirms a fact helps: a land title or a vineyard lease, papers for the production site, registration data, a dated page on your own website. An objection without evidence will be read, but it cannot take the place of a fact.
A submission becomes its own task in the review queue — with the producer, the contact, the substance of the objection and a note on the documents attached. From there a person handles it: checks the facts, records or corrects statements with their sources, and recomputes the classification.
The category never moves automatically, in either direction. A person confirms the decision, and it stays in the classification history for good: the previous category, the date of the revision and the methodology version are all kept.
There is no submission form on the site yet, and we are not going to invent a review deadline we cannot guarantee. The rule holds regardless: a submission with documents is handled by a person, and the outcome — a corrected fact, a revised category or a reasoned refusal — is recorded in the database.
If the disagreement is not about a fact but about the methodology itself — the weight of a competition, say, or a threshold — that is a legitimate conversation too. The weights and thresholds are published and live in the code: they are changed by review and tests, not on individual request.
Data and corrections
The data comes in four layers: public registries (appellations and their registration dates follow the Sakpatenti register), producers' own sites and labels, shop catalogues, and the official winners' lists of competitions. From shops we take the price, availability, the name of the listing and the link.
Different data moves at different rhythms. Prices are collected in a full weekly pass; once a week we also look for changes — listings that appeared or vanished in shops, winery pages that were edited; producers' sites are crawled monthly. The pace is deliberately gentle, over public pages, without working around any restriction.
Price observations are written once and kept: every price stays in the database with its own date and shop. History therefore accumulates from the very first run, and a wine catalogued later inherits all of it instead of starting from zero.
Part of the work is automatic. Linking a source record to a catalogue record is applied on its own once the confidence computed from weighted signals passes a threshold; where the match is ambiguous nothing is decided at all, and anything decided can be undone. That relaxation does not extend to classification or to publishing facts about companies — there the decision stays with a person.
The database is assembled from open sources and it contains errors — that follows from how it is built, not from carelessness. We would rather show «no data» than fill a gap with a guess, so an empty spot on a card is itself a piece of information.
Found a mistake — tell us. A correction that names its source is the fastest kind to verify, and the handling is the same as for the right of reply described in the section above.