Hacker News Reader

Comment by clnq | Hacker News Reader

I always wondered why DarkSky was so popular. Was it the design? It was known to be generally inaccurate among meteorologists, while other apps (METAR+TAF especially) would be remarkably accurate.

I don't intend this to be a jab at anyone's app preferences, but I'm very interested in the value proposition that hooked people.

View on HN

In rainy parts of the US such as the Pacific North West coast, having a thirty minute warning that it's about to rain is really handy. It can make the difference between biking home sopping wet or squeezing in between two downbursts of rain. The rest of the app was fine, but not great. It demanded constant attention as its "predictions" changed constantly.

This! It was that combination of rain radar and really obvious app specific audio alarms that no other weather app seems to understand.

I live in Scotland, a famously rainy country. We line dry our laundry outside and dark sky alerted me take it indoors before rain for years.

Then Apple took it away and my laundry regularly gets rained on.

The closest I've found for this functionality on Android is Rain Alarm. It has two flaws though. One is that it uses public rain radar data that only updates every 15 mins, so often you get an alert when its already started raining. The second is that it uses a circular search path, so you get false positive rain alerts that it's raining 3km away when that rain shower is actually passing by and will not hit us.

For these features Dark Sky was unique and accurate enough.

I've had decent luck with the Met Office app for Android

The weather is only at 1hr intervals but the rain radar maps are at 15min and you get the last 6 hours and the next 24, so you can see where patches of rain are coming from and where they're going to be at which time

At least for me here in the UK the live-updated rain was very accurate, you could watch it come and go in pretty much real time combined with looking out the window.

I'm struggling to find anything like it now, it was incredibly useful since I go out a lot to do little jobs and if its a crappy weather day (pretty much every day this time of year) it was nice to check to see if I had enough time to do something or if I'd need to go out and cover some things up.

I've never used the app, only the website.

I find rain alarm good enough for this.

Most users aren't meteorologists. I found the precip radar view good and it was smooth for projecting out a couple hours which was important for deciding if I wanted to golf or not because it wad easy to see if a storm was likely to clip my club. That's something none of the current apps I've tried since replicate nearly as well.

They were effectively first easy-to-use near-/short-term forecasting service.

It was also a true "indie" project with 2 (3?) guys with no relevant background deciding to tackle an annoyance in their own lives. They did it and it happened to work well. They did a kickstarter. Then they tried being aggressive with the growth and it didn't go so well.

Also their dataviz was really pretty. That helped.

I always found that for very near term forecasts it was better than alternatives. But the further out my time horizon the less good it was.

“What’s this weeks weather going to be” I’d use another source.

“Will I get rained in if I leave the house now, or should I wait 30 mins?” I’d use dark sky

I’ve switched to Apple Weather which has been a reasonable replacement but I don’t like it as much

Apple claimed to have integrated Dark Sky's technologies into their Weather app but clearly haven't. I've relied on Dark Sky's rain notifications for years and they've always been remarkably accurate (UK here), but the stock Weather app sends maybe one notification per 100 rains. So far this year, since Apple turned off Dark Sky, I haven't received a single rain notification and consequently I've been absolutely drenched on several dog walks.

I ran both in parallel for a while. Local to me they were pretty close but not quite the same. Dark Sky was always more accurate when they differed.

Reply on news.ycombinator.com