How we reach a verdict

The idea the whole product rests on

Why we don't just show you a forecast

A generic weather app reduces the sky to one number: percentage cloud cover. A forecast reading "90% cloud" sounds like a wasted day. It might be — or it might be the best day of the month.

What matters at altitude isn't how much cloud there is. It's where the bottom of it sits relative to the summit. A low deck of cloud can sit entirely below a viewing platform: from the valley the sky looks like a solid grey lid; from the summit, you're standing in bright sun above an unbroken sea of white, with only the tips of other peaks breaking the surface. That is the single best kind of day this site exists to find, and it is a day every weather app calls "cloudy".

The reverse is also true, and it's the case a bare "cloud cover" percentage can't tell apart from the first: a deck whose base sits right at the summit's elevation. From the valley, that also reads as "90% cloud". From the summit, you're standing inside fog, seeing nothing at all.

We only know where a cloud layer begins — the forecast model reports a base, not a ceiling — so a base only slightly below the summit is treated as a real risk of standing inside cloud, not a promise of clear air. A margin of at least 500 m is what separates a trustworthy PEAK from a lucky guess; 800 m or more is the dramatic sea-of-cloud day.

There's a fourth, quieter case worth knowing about too: sometimes the deck sits entirely above the summit. You get clear air at the top, just under a grey sky rather than a dramatic view down onto cloud — a reasonable day, not a spectacular one, and we score it as such. And most good days aren't dramatic at all — they're simply open sky, with less than a quarter of it showing low cloud. That starts a strong score on its own, trimmed slightly for hazier mid- and high-altitude cloud that dims the light without hiding the view.

This is why our verdict and your phone's forecast will sometimes disagree on the same mountain, on the same day — and why we show you the geometry instead of a summary percentage.

Stage by stage

What goes into a verdict

The score behind every verdict starts from whether there's a view at all, then subtracts for everything that would spoil the day. View quality dominates, because the view is what you're actually buying.

View quality comes first

We start from the cloud geometry above. Below that, horizontal visibility — how far you can actually see distant peaks — subtracts further: nothing at 20 km or more of visibility, down to a heavy penalty below 2 km. Visibility comes from a secondary weather model rather than our primary one, so when it's the main thing pulling a verdict down, we say so and lower our confidence too — more on that below. And if we can see a deck of low cloud but the model doesn't report where its base actually is, we don't guess the geometry — that hour scores a flat 50 and the forecast is marked uncertain.

Then precipitation

Rain and snow subtract from there. Any meaningful rain — 0.5 mm/h or more — costs real points; 2.0 mm/h or more is a heavy penalty, because rain ruins both the view and the visit. Snowfall of 1 cm/h or more gets its own penalty on top: it's beautiful, and it still hides the panorama you paid for.

And wind — counted twice, for two different reasons

Wind appears in the score twice, deliberately, because it means two different things and collapsing them into one number would be wrong in both directions.

The first is operational risk: how much wind it takes to stop the machinery. An aerial system stops for wind long before a ground-based one does, so we track gusts against different thresholds by transport type:

High operational risk caps the verdict at MAYBE — never SKIP on wind risk alone, because we don't actually know the operator has stopped running, only that they might. Claiming a closure we haven't verified would be exactly the kind of guess this site exists to avoid.

The second is comfort, and it's deliberately mild: people accept being cold and blown around at altitude — they're there for the view. Gusts of 60 km/h or more, or an apparent temperature at or below −15 °C, cost a few points. Not a verdict, just an honest nudge to dress for it.

The gate nothing else overrides

Operations are a hard gate

None of the above matters if the mountain is closed. A closed cable car, cog railway or funicular can never produce a PEAK — however good the sky looks, if you can't get up there, there's no view to have.

Here's the part we'd rather not have to admit, but it's the honest state of things: we currently know an operator's published schedule, not whether they're actually running right now. Outside opening hours, we can say "closed" with confidence. Inside opening hours, our best honest answer is "unconfirmed" — and that deliberately lowers our confidence rating, rather than letting the page quietly assume the lift is running. Live operator status is on our roadmap. Scraping operator websites to fake it in the meantime is not something we're willing to do.

Not just "today"

How we pick the best window

A verdict is for a specific window, not a vague "today". We only ever consider hours a visitor could actually be up there: the attraction's own published service hours for that date, narrowed further to between sunrise and sunset. A beautiful forecast at 3 a.m. is not a reason to go anywhere.

Those hours are the operator's, not a rule of ours, so they differ by mountain and by season — Harder Kulm's funicular runs into the evening in high summer, and a winter timetable closes earlier than the same mountain's summer one. We used to apply one fixed 08:00–17:00 band to every attraction, which both hid genuinely good evening hours and recommended hours that were already shut. That band now only applies to an attraction whose schedule we don't hold.

Within those hours, we look for the best contiguous 3-hour block — shorter if the day genuinely doesn't offer a full three hours of decent conditions — and score the day on that window's average, not the day's average. If the morning is glorious and the afternoon clouds over, averaging the two would call the day mediocre, when the honest advice is: go in the morning. We'd rather tell you exactly when to go than hand you a number that blurs a great window into a forgettable one.

No black box

The thresholds, in the open

The final score, 0 to 100, becomes a verdict by two lines:

• score ≥ 75 → PEAK
• score ≥ 45 → MAYBE
• otherwise → SKIP

PEAK is set deliberately high. A PEAK is a promise — go, spend the money, spend the day — and a PEAK that disappoints costs us far more trust than a cautious MAYBE ever earns. We'd rather undersell a good day than oversell a mediocre one.

Two hard rules override the number entirely, and both are described above: operations closed or scheduled-closed forces SKIP, score 0, regardless of the weather; high wind operational risk caps the verdict at MAYBE, whatever the score says.

How much to trust the number

Confidence

Every verdict also carries a confidence rating — high, medium or low — separate from the score itself. It starts at high and only ever gets downgraded, never upgraded, by the single worst thing that applies:

• A forecast lead time of 2–3 days out, or 4 or more days out for a bigger downgrade — forecasts get less reliable the further out they are.
• Weather data older than 90 minutes, or older than 3 hours for a bigger downgrade.
• An operating status we can't confirm live — see "Operations" above.
• A cloud deck we can see but can't measure the base of — we won't guess the geometry.
• Visibility doing most of the work in a low score — it's our weakest input, and leaning on it should be visible.
• A score sitting within 4 points of a verdict boundary. A 74 and a 76 are the same forecast, and pretending a line at exactly 75 is more precise than that would be false confidence.

Feature, not apology

What we don't know yet

This is v1, and the thresholds above are informed judgement, not yet empirically calibrated — we've set them where the physics and the stakes argue they should sit, not where a dataset told us to. Being honest about that means being honest about a few specific gaps:

• We know an operator's published schedule, not live running status. Real-time confirmation is on the roadmap.
• We don't yet check any webcam against our own call. If and when we do, frames will only ever come from cameras we have explicit permission to use — never scraped or re-hosted without it.
• We run one weather model. When we add a second, disagreement between them becomes a confidence signal in its own right — right now we have no way to know when our one source is simply wrong.

We record every forecast we see and every verdict we issue, on a fixed schedule, starting from launch. That's not for later marketing — it's so that, in time, we can compare what we predicted against what actually happened and publish the result: how often a PEAK was actually worth it. As far as we know, nobody else in this space publishes their reasoning, and nobody publishes their accuracy either. We intend to, once there's enough data to say something honest.

Who the number answers to

Independence

The part of this system that decides PEAK, MAYBE or SKIP cannot see affiliate data. It has no import path to partner IDs, pricing or booking links — that's not a policy, it's enforced by an automated test that fails the build if it's ever violated. Offers are attached to the page only after the verdict already exists, by a separate part of the codebase, and they never reorder or influence which mountain we recommend.

That matters because a SKIP verdict costs us money — nobody books a ticket off a page that just told them to save their francs and try another day — and we publish it anyway. If we were optimising for revenue, every day would be a MAYBE.