AccuWeather Personalizes Forecasts on a Global Scale

Digital Media VP David Mitchell explains how the company localized its mobile apps in 60+ languages.
Customer Experience

Delivering weather forecasts across the globe includes numerous components from language translation to localized content and terms. A surfer in New Zealand, for example, probably cares more about wind speeds than an office worker in Germany. As AccuWeather grows its global reach, it's localizing forecasts to improve the user experience.

1to1 Media spoke with David Mitchell, vice president of Digital Media at AccuWeather; and Ian Henderson, chairman and chief technology officer at Rubric, about personalizing weather information for a global and digital audience.

1to1 Media: What experience do you want people to have when they go to AccuWeather's site or app?

David Mitchell: We want to personalize the weather to improve our customers' lives. It's an ongoing effort and that includes translations of local reports. We're focused on our global audience and reaching them in the ways they want to consume weather information. The way people consume weather information is rapidly changing. We're seeing this transition from TV forecasts to online and increasingly mobile. The bulk of our traffic occurs on mobile devices. We're continuously looking at how we can make the forecasts we provide more personal as device capabilities and contextual information grows.

When did your partnership with Rubric start and how do they help you personalize your forecasts?

DM: We have relationships with handset manufacturers to have AccuWeather preloaded on phones, which has dramatically increased our audience outside of the U.S. Our relationship with Rubric started about two years ago when we realized some of our language translations were deficient. Rubric focused on what our content is, where the context can be skewed, and if it feels local by ensuring we were using terms that a local would use. We continue to get better as the relationship progresses.

Can you give me an example of a personalized local forecast?

Ian Henderson:
Customers can get forecasts in their own language. When we started working with AccuWeather, we were doing translations for 40 languages and now it's up to 60 languages and growing. For example, AccuWeather has forecasts in French for Canadians and also French for people in France. It also has forecasts in Spanish for Spain and individual Latin markets as well. Particularly in a mobile environment, customers expect to get information in their natural language and their dialect.

We're also working with AccuWeather to pin down what people want in different environments. For instance, in the U.K. the pollen count is very important, but not as much in Egypt. We've heard pollution levels are a big issue in Beijing, but not in South Africa. And if you're in New Zealand, you want to know long you can be in the sun before you burn.

DM: We're a U.S.-based company but more than half of our traffic comes from outside the U.S. across all our digital media properties. We've made a significant amount of progress in localizing information, but there's still work to do.

What factors do you use to identify the correct language for each customer?

DM: What we're mainly looking at is the language setting on your device and the location of the weather information that you want.

IH: Say I'm in France and I look up I'm automatically sent to but I barely understand French. This is where AccuWeather comes in. It looks at the device's setting and can detect the language and region that you prefer.

How do you scale your customized reports?

DM: To be honest, that has yet to be determined, but it helps that the serving platform for our websites and apps is heavily cloud-based. Scaling is relatively easy at this point. We do very little endpoint delivery to devices since we use cloud services all around the globe. Six months ago our API was handling six billion data requests per day and now it's getting 9.5 billion data requests per day just to give you some context.

Deciding how to store this information and get it right each time is also a work in progress. That involves questions like, "How much of the data do we want to be cookie-based or location-based?" We could potentially set up cookies based on user behavior and eventually use logins to get a more robust profile of a user.

What KPIs did you use to measure the impact of the translated reports?

In the beginning, we were seeing really high bounce rates from our global customers. So we looked to see if improving translations for particular regions would lower the bounce rate. We also look at the number of pages consumed, what pages those are by locale, how often users come back, and to some degree, time spent on site.

How else does AccuWeather personalize its interactions with users?

DM: The biggest part of my focus is emerging platforms and the Internet of things including wearables, the connected home, and connected car. We're working on ways to match our information with the data that's available from all these connected devices. So, if we know a user has a Garmin smartwatch [that's paired with a smartphone or apps] to work out with and we know he's a runner or a swimmer, we'll send forecasts that are more useful to that person. It all comes back to how we provide relevant information based on location, time, and context to individuals.