Our April release ships three changes our customers have been asking for: rank shown everywhere it matters, a single home for data exports with presets and history, and a new API endpoint that exposes the ground truth we score every provider against. Here’s what’s new and why it matters.

Positional rank, on every screen

Forecast providers don’t always ask “what’s our percentile?” They ask “what’s our rank?” Rank is the number that ends up in marketing claims, in board reviews, and in internal OKRs — and within the new platform, until this release, it was the one number ForecastWatch made you compute in your head everywhere except the leaderboard view.

We’ve fixed that. Ordinal rank now appears as X of N peers directly on the cards and tables you actually look at: hero cards on Monthly Insights, metric tiles on Market Overview, and a new Position column on the Performance Comparison table. Same number, same place, on every screen. Snapshot PDFs picked up the same treatment alongside a small design refresh, so an exported Market Overview now drops cleanly into a board deck without a re-skin.

A small change with an outsized practical effect:

• Faster review cycles. “How are we doing in EU temp this month?” answers itself the moment the page loads.
• One number that travels. The rank on the card is the rank on the PDF is the rank in the comparison table — analysts stop disagreeing about which translation of percentile to use.

Rank is computed on the exact metric we publish, so the number on the card is the number you can quote.

One Data Export hub, with presets and history

Every weather company has someone whose Monday morning starts with the same export: last week’s accuracy by station, by region, by lead time. CSV. Same five filters. Every Monday.

We’ve replaced the scattered, per-screen export modals with a single redesigned hub at /export. Every export button — Monthly Insights, Daily Insights, Long-Term Trends, Forecast Risks, Station Metadata, Station Watchlist, and the new Forecast Watchlist — now opens the same page with your filters carried over, plus a one-click link back to where you came from.

Two new tabs do the real work:

• Saved Presets. Name an export configuration, save it, and re-run it next week with a single click. Presets are personal to your account, and each one stays tied to the provider context it was created in — so re-running is always against the right peer set.
• Export History. Your last 50 exports are auto-captured and re-runnable. An audit trail and a re-run button in the same panel.

For QA leads, a recurring weekly chore drops from minutes of filter-resetting to a single click, with a record of what was pulled. For analysts doing one-off investigations, it’s a way to revisit a slice you already explored without remembering exactly which filter combination produced it.

If you’ve ever rebuilt the same export by hand because you couldn’t remember the configuration from last quarter, this release is for you.

A new API endpoint for canonical ground truth

Anyone who has tried to score forecasts against ground truth knows the first hard question isn’t how — it’s which. Pull observations from one source with default parameters and you get one stream; flip a flag and you get a different, canonical record. The two disagree at the edges. And if your internal QA pipeline reads one while ForecastWatch is scoring against the other, your numbers will quietly disagree with ours — and the disagreement won’t be about math, it’ll be about data sourcing nobody documented.

This release adds a new endpoint to our v1 API to remove that ambiguity:

GET /v1/observations/?station_code=KORD&start_date=2026-04-01&end_date=2026-04-07

What you get back is the exact dataset ForecastWatch uses to score every provider in the public benchmark — daily ground truth, by station, up to 366 days per request, in your choice of imperial, metric, or meteorological units.

A few details worth flagging:

• Lookup by code or ID. station_code accepts ICAO (KORD) or SYNOP — whichever your internal systems use.
• Sensible range cap. A year-and-one-day per request keeps queries fast and predictable. Stitch larger ranges client-side if you need more.
• Three unit systems out of the box. Imperial, metric, and meteorological — no client-side conversion.
• Same authentication as the rest of the v1 API. Bearer-token auth, scoped to your account.

Until now, our v1 API has exposed computed metrics — insights, market position, risks, watchlists — built on top of the scoring layer. GET /v1/observations/ is the first endpoint to expose the layer underneath: the raw, authoritative ground truth that every score is computed against. If you’re building an internal QA pipeline, a dashboard, or a post-hoc validation harness, this is the hook you’ve been asking for. And if you’ve ever lost an afternoon reconciling your accuracy numbers with ours, it’s the endpoint that ensures you never have to again.

Login in now to see for yourself!