The forecast is the brief: weather-based advertising that launches before the rain does

Most weather-based advertising is a scramble. The rain starts, someone in the office notices, a campaign gets duplicated, the copy gets rewritten, and the ads go live around lunchtime. By then the people who needed an umbrella already bought one somewhere else. Reacting to weather is easy. It is also late.

We built weather-based advertising the other way around: forward-looking. The system reads the forecast the way a media planner would if a media planner had infinite patience, and it treats the forecast as a brief. When a condition worth advertising against shows up in the data, the ads exist before the weather does.

How does weather-based advertising actually work?

Our system watches 7 weather types up to 10 days ahead, hour by hour, in every city the business serves. Each city gets its own reading, because a cold front is a local event and the ad about it should be one too. A regional average would smear a storm across places it never touches.

When a condition crosses its trigger, the system builds ads for exactly the cities in the path of that condition. It writes the creative angle for the weather in question, schedules the launch for the moment the ads should start earning, and, this is the part almost everyone misses, writes the end date at the moment of creation. The ads retire themselves the day after the weather passes. Nobody has to remember to pause anything. There is nothing sadder than a "rain is coming" ad running under a blue sky. The way to avoid that is to write the exit into the ad on the day it is born, before anyone gets a chance to forget.

The night rain crossed 55%

A real night, recently. The forecast updated overnight and rain probability crossed 55% for three cities a client serves. The trigger fired. The system built ads for those three cities, only those three, scheduled the launch, stamped the end date. The ads existed before anyone woke up. The client found them in the morning the way you find bread the baker made overnight: finished, warm, nobody had to ask.

That is the whole argument for automation here. Weather keeps its own schedule and forecasts update while you sleep. A human can check the forecast every morning. A system checks it hour by hour and never decides the couch looks more appealing. If you want the version of this where the ads build themselves while you sleep, that is literally what happened that night.

Should weather ads run in the middle of the night?

Here is the nuance that separates a good weather trigger from a noisy one. Cold in the small hours does not become an ad for a store that opens at 8am. Nobody drives out to buy a heater before dawn, and the coldest hour of the day is usually the least commercial one.

So the reading respects the hours when weather actually matters for that business. The default window is 8am to 8pm: the system evaluates the forecast inside those hours and ignores the theatrical extremes outside them. A city can hit its coldest hour long before sunrise and trigger nothing, because by the time the doors open the day is merely cool. The opposite holds too. A modest afternoon condition inside the window can matter more than a dramatic one outside it.

There was no approval step in that story, on purpose. A forecast window is short, and an ad waiting for a yes can miss the rain entirely. This is the case where executing directly is not a convenience but the feature itself: the system builds, schedules and retires the ads by itself, and the log tells you about it in the morning.

Ads that know when to leave

The end date deserves its own section because it is the piece almost everyone skips. Most weather campaigns are born without a death date, which means they die by neglect: someone notices weeks later, usually while auditing spend. Ours are born with the exit already written. The day after the weather passes, the ads retire on their own and the account returns to its normal state without a meeting.

Forward-looking, scoped to the exact cities in the path, with a scheduled entrance and a guaranteed exit. That is weather-based advertising as we practice it inside Attribution. The forecast was always public data. The difference is having something awake enough to act on it.