As I write this essay, today’s weather forecast is for a high of 75 and a low of 61. I have no idea what that means.
It’s going to be nice, for sure, but will it be 75 this morning or sometime later in the day? Will it be 75 for a while or only a brief moment? 75 to 61 is a 14-degree range, so is the spread dramatic or not? Does 75 feel almost like 80, or more like 70? Something in the low 60s isn’t necessarily warm but coolish, right?
The humidity also changes everything, so the same temps could feel very different if there’s lots of moisture in the air. And whatever the day feels like to me, it’ll feel differently to the next person because our senses are unique from one another’s, at least in part. Our reactions to objectively similar experience as different, too (i.e. one person’s grossly hot day is another’s nirvana).
I don’t know about you, but I find most of what I get in even the best weather forecasts to be pretty much useless. No amount of excruciatingly detailed information makes things any clearer. It’s like a lot of color commentary on a sporting event for which I don’t understand the fundamental rules of the game.
I have a simple solution to the conundrum: start every forecast with a recap of the prior day’s weather highlights. Skip reporting highs and lows, and instead report the mode (or at least an average). And come up with a better method for calculating probability, and apply it to the temps in addition to expectations of storms.
Call it a weather recast instead of a forecast. Here’s my reasoning:
- I have a tangibly explicit recollection of what yesterday felt like to me, so it’s a foundation on which I can better comprehend today’s forecast. I almost don’t even need the numbers. Just tell me if today will be better or worse. Adding all of the whys, from cold air fronts moving across the Rockies to butterflies flapping their wings of Beijing, is fine additional content, but the only way I understand my future is with some reference to my past.
- Instead of bookending the temperature, tell me what temp will be the most common throughout the day (in math, this is called the mode). I don’t care if the high will happen for a nanosecond, or that it will be really cool while I’m asleep. Four hours of so-and-so temp makes far more sense to me. An average temp would work, only not as well.
- Anybody understands probability (it’s why Vegas and workout gurus make money), so I’m not embarrassed to admit that I know only that the chance of a storm at 40% vs. 30% is 10% more of…something. Why not correlate those calcs with historical data for both same day and similar conditions, then translate it into a clearer, perhaps even binary result? We’re all gambling on the weather, and forecasters risk being wrong even after they couch their insights into qualified gobblygook. Tell us yes or no.
Such an approach suggests some interesting ideas about how we look at marketing communications, too:
- Do our promises of product or service experience adequately reference consumers’ known experience, so they can make more reasoned decisions?
- Do our functional benefits — from processor speeds to freshness — stand as exceptional, idealized qualities, or do they represent the most commonly-experienced attributes (and does anybody understand the difference between a gig of memory or a “sell by” date on a carton of milk)?
- Do our statements about reliability, such as J.D. Power rankings, reflect the expectations consumers should have of actual performance? Like the weather, one person’s dissatisfied ownership can be another’s minor blip in the road, so to speak.
Well, I’m heading out now. Still not sure if I need a jacket.
(Image credit: weather.com)
Original Post: http://www.dimbulb.net/my_weblog/2011/05/a-better-forecast.html