In 2012, The Weather Company (TWC) was at a crossroads. Founded in 1982 as “The Weather Channel,” the company’s core business centered on cable television programming, including weather forecasts and other weather-related content such as documentaries. But for years, viewership had been declining in the face of “cord-cutting” by consumers, as well as the proliferation of digital weather tools made ubiquitous by the smartphone. Yet, TWC had no mobile weather app—indeed, no real online presence to speak of.
Confronting these headwinds, the company embarked on an ambitious plan to leverage big data to propel the company into new, digital-first business verticals. The company announced its pivot by rebranding as “The Weather Company.” The embrace of data analytics was a natural next step despite the company’s decades-long roots in the cable television business. Said CTO Bryson Koelher: “Weather is the original big data application… when mainframes first came about, one of the first applications was a weather forecasting model.”
The company’s digital transformation entailed three main steps:
First, TWC augmented its deep bench of talent in meteorology and atmospheric physics by hiring data scientists and machine learning experts to work on cross-disciplinary teams building more complex forecasting models integrating a wider array of data. Relatedly, it expanded its pool of weather data by acquiring Weather Underground, which offered a weather app that used “crowd-sourced microclimate data” to offer precise weather forecasts through its mobile app.
Second, TWC overhauled its IT infrastructure by moving its 13 data centers into the cloud and integrating the “loose-knit collection of aging applications” constituting its legacy platform into a single integrated data platform. This integration not only enabled more efficient data analytics processing, but it also made it easier for third-party partners to plug into the platform. Costs for each API call—when a third-party application called on TWC servers to access weather data—plummeted to 1/70th of the cost under the pre-cloud infrastructure, driving customer growth. Today, TWC is one of the world’s largest API platforms, with over 25,000 partners and over 26 billion daily API calls.
Third, TWC began a period of experimentation to understand new verticals where its big data strategy would generate value for its business customers. A key vertical that emerged was marketing and advertising. TWC hoped to generate hyper-localized insights about consumers’ behavior by analyzing “when, where and how often people check the weather,” which would allow businesses to better target their advertisements.
TWC’s early experiments yielded surprising findings, according to a Wall Street Journal report:
- Consumers in different locales respond to the same weather differently:
- Bug spray sales increased during spring in Dallas when the dew point was below average, whereas bug spray sold better in Boston when the dew point was above average.
- In Chicago, heatwaves caused a surge in air conditioner purchases within a day, whereas consumers in Atlanta—ostensibly more acclimated to heat—tended to wait more than 2 days during heatwaves before purchasing relief.
- When a consumer checked the weather affected their consumption habits:
- Michaels’ craft stores had hypothesized that sales would surge on rainy days—when many people would take up craft projects. Instead, TWC insights showed that Michaels’ store sales surged when weather forecasts predicted rain 3 days in advance.
- Individuals who checked the weather earlier in the day were more likely to make decisions about how to plan the rest of their day—including potential shopping trips.
In November 2015, TWC was acquired by IBM. In November 2019, IBM launched Global High-Resolution Atmospheric Forecasting (GRAF), which brought IBM’s supercomputing power together with TWC’s weather prediction know-how. GRAF promised forecasts that refreshed hourly on areas as small as 2 miles wide (as compared to every 6-12 hours, on areas as small as 6-9 miles wide, under the most advanced weather forecasting alternatives then on the market). The supercomputer underlying GRAF claimed to process 12 trillion pieces of weather data integrated from an array of sources including smartphones and airplanes. GRAF promised to continue extending TWC’s pioneering approach to weather-related predictive analytics into new applications across a range of industries.