Asking around and researching features for the unnamed temperature app, something that kept coming up was the idea of being able to see yesterdays temperature.
As Neven Mrgan articulated expertly in a rant on the native OS X weather app:
“…And it would make room for a feature I’d really like to see in every weather widget, on the phone or TV or wherever: yesterday’s weather. The point would be to give me some indication of how much cooler/warmer I should expect to feel as compared to yesterday. See, I don’t really know how to judge absolute temperature values. Which coat do I grab at 55º? How about 43º? If, however, I knew that it would be 8º cooler than the previous day, I could act appropriately.”
We wondered why more weather apps hadn’t implemented historical data like this, as it’s hard to argue about the contextual benefits. Well, as we would soon find out, including historical data isn’t that straightforward a deal.
First of all, finding a source with reliable world-wide historical data of yesterdays weather was surprisingly hard. We’re currently in contact with some subscription based services that will be able to supply us with the data for a fee. We’d like to avoid setting up our own server so this looks like the only option right now.
Secondly, and probably the biggest conundrum was, what exactly do you display as yesterdays temperature? The average temperature of the entire day? The median? The temperature at the same particular time yesterday? Using something like the average temperature wouldn’t give you a very good idea about temperature you might have experienced, as temperature obviously varies a great deal from day to night making any attempts at summing up a single number a muddy affair.
If we’re able to get temperatures from approximately the same time yesterday as the ones the user is being given currently, that would definitely serve as a better ground for comparison. The problem with this approach, is that yesterdays temperature, a number the user might believe to be a fixed number, would appear to inexplicably change during the day when the user refreshes the data at different times.
There’s problems accompanying both solutions, the average of all of yesterday seems too inaccurate to be useful when compared to a snapshot of todays temperature. Serving up comparable temperatures within the same timeframe in the two days, makes the most sense when you’re trying to get a contextual reference - but it might be confusing to see yesterdays temperature jump up and down, and depending on the data we can get, it might not even be possible to get that amount of detail.
Is it even useful? If you got any bright ideas you’re welcome to shoot as an email or hit me up on twitter
1 note