Five levels, two corrections, and what we still get wrong.
Twelve of the thirteen cellar apps we have compared say they tell you when to drink your wine. Almost none of them say how. That matters, because the methods behind that promise are not comparable, and the difference shows up on the bottles you care most about. Here is how ours works, and where it is still weak.
Caviste starts from the most precise thing it knows about your bottle and falls back level by level: estate, cuvée level, appellation, region, grape. Two corrections then apply to your bottle rather than to its origin: the vintage and the format. Nothing falls through, every window stays editable, and we measure our own error instead of asking you to rate us.
Three approaches exist in this market, and they fail in different places. Reading the producer's own recommendation is the most authoritative answer there is, until the producer says nothing, which is most of the time for a small grower. Interpolating grape ageing curves covers everything, at the cost of treating a first growth and a supermarket red from the same grape as the same wine. Leaning on a wine database is precise for the wines that are in it and silent for the bottle you brought back from a holiday.
Ours is a table, resolved from precise to general, that never runs out of an answer. Its weakness is the opposite of a database's: it is only as good as the rules we wrote. Which is why most of this page is about how we find the wrong ones.
Each level is tried in turn, and the first one that answers wins.
| What it knows | Entries | |
|---|---|---|
| 1. Estate | That this particular producer outlives its appellation, and that a second wine is not the grand vin | 217 estates |
| 2. Cuvée level | Champagne only: whether the bottle is non-vintage, vintage or a prestige cuvée | non-vintage, vintage, prestige |
| 3. Appellation | The ageing potential of the named appellation, label mentions included | 855 appellations |
| 4. Region | The regional average by colour, when the appellation is unknown to us | 225 regions, 40 countries |
| 5. Grape | Last resort: the backbone grape decides, because a blend keeps as long as its most solid variety | 74 varieties |
An appellation rule describes the average producer. Pauillac keeps thirty years is right for a thousand châteaux and wrong for five of them, in both directions: the celebrated ones outlive it by decades, and a second wine from the same estate never reaches it. Named estates are therefore checked first, and when we have no rule for one, the answer falls back to its appellation, which stays honest.
The table makes the appellation carry the rank of the wine. That works in Burgundy, where the hierarchy is in the name (village, 1er cru, grand cru). It does not work in Champagne, where the whole region is a single appellation: a prestige cuvée from a declared year and a brut with no year come out of the same line, when the market gives one fifteen years and the other three. Champagne goes through a cuvée level instead, and those levels are published on the Champagne page.
Some mentions legally impose a longer ageing before release, and the wine that carries them keeps longer: a 1er Cru outlives its village, a Gran Reserva outlives a Crianza. Caviste reads them from the appellation you typed. Others, like Grand Cru or Classico, designate a zone or a rank the table already carries in its own lines, so they change nothing on their own: counting the same thing twice would be worse than ignoring it.
643 rated years across 24 regions, on a deliberately coarse five-notch scale from exceptional to weak. We do not pretend to separate a 92-point year from a 94-point one, and a year we have not rated is neutral rather than guessed at. Colour matters as much as region: the sweet wines of Bordeaux do not share the great years of its reds, 2007 being mediocre in red and immense in Sauternes. A great year lengthens the window and a weak one shortens it, including for the longest-lived wines. That is what separates a legendary 1961 from a washed-out 1974 at the very same address. Year by year, the windows are published in the vintage charts.
Wine ages more slowly in a large bottle and faster in a small one, so a magnum stretches the window and a half bottle shortens it.
From the vintage, normally. A sparkling wine with no vintage used to get no window at all, and therefore no gauge, no reminder and no place in the list of what to drink. It is now counted from the year you bought it, the only date that means anything for a non-vintage bottle. Elsewhere, a missing vintage is an incomplete entry rather than a property of the wine, so we leave it alone.
Worth stating plainly, because a method you cannot criticise is a marketing claim.
This is the part no comparison table has a column for. Every app can claim drinking windows; the question is whether anyone ever finds out that theirs are off.
We measure it, and we ask you nothing to do it. When you rewrite a window we suggested, Caviste compares your years with what the rule says today and keeps the gap. Those gaps are averaged per appellation, because that is the form that turns into work: one person disagreeing tells us nothing, while several people pushing the end of the same appellation back by years points at a line to reopen.
Not adding a rating button was deliberate. Caviste already offers a thumbs up elsewhere and it is barely used. Voting is a free gesture you make for other people; correcting is one you make for yourself, and only the second one reliably happens.
We publish no accuracy figure here yet. The measurement started in August 2026 and the corpus needs weeks of ordinary use before an average means anything. We would rather leave this paragraph empty than fill it with a number we cannot stand behind, and the figures will go here when they exist.
A wine past its window is not a wine to throw away, and an app that turns it red and sounds an alarm is an app you learn to ignore. Caviste says window passed, tells you on the bottle sheet how many years ago that was, and leaves a long-ageing wine out of the urgent reminders. The judgment stays yours: you are the one who knows how the bottle was stored.
Try Caviste for free. 30 bottles, 50 AI scans, no card. On the App Store and Google Play. See the app →
It depends on four things, in this order of weight: who made it, where it comes from, the vintage, and the bottle format. Caviste resolves them in that order, from named estates down to grape varieties, so it always has an answer. A Champagne with no vintage keeps a few years, while a prestige cuvée from a declared year keeps decades. These are indicative oenological benchmarks, and every window stays editable on every bottle.
Because the appellation is not the only thing that decides. An appellation rule describes the average producer, so Caviste checks named estates first and reads the label mentions that legally impose a longer ageing. The vintage and the bottle format then move the result: a magnum stretches the window, a half bottle shortens it, and a weak year shortens everything.
We measure them. When you rewrite a window we suggested, Caviste compares your years with what the rule says today and keeps the gap, then averages those gaps per appellation. Several people pushing the end of the same appellation back by years points at a line of the table to reopen. We ask for no vote and no rating: the measurement runs on corrections people make for themselves. The corpus is young, so we publish no accuracy figure yet rather than invent one.
See also: Ageing potential by appellation · Vintage charts · All the cellar apps compared · FAQ