How ArabiaWeather leveraged big data and processing power to stay ahead of the competition
Hurricane approaches. (Image via symonsez.wordpress.com)
The story of ArabiaWeather is one referred to in the Middle East startup community again and again - so often that it's almost become legend. Founder Mohammed al-Shaker started the weather prediction service when he was still in high school back in 2006, and since then the site has since grown into one of the largest content sites in the region. As of last month, the mobile app boasted 289,000 active users, 3.1 million unique users visited the website, and the site was ranked number one in the MENA in the travel category, and number five in the MENA in the portals category.
Part of its success, says Chief Product Officer Yousef Wadi on a call with Wamda, is that ArabiaWeather has leveraged existing, largely Europe- and US-specific weather algorithms to create an accurate and nuanced prediction mechanism, as we have covered previously. This has been a rapidly accelerating process over the course of the company’s nearly 10-year journey (ArabiaWeather has been operational for four years, having transitioned from a Jordan-specific site).
Processed data and supercomputers
As the team gears up for its biggest year yet (which will include the launch of several new products and services), Wamda asked Wadi what other technologies the startup has adopted that have impacted its formula, as well as how the product is sold.
When the team made the shift from JordanWeather.jo to a region-oriented service, the way they collected and processed data had to change due to the sheer amount of information they were now dealing with. “The amount of data generated by these models [the weather algorithms as well as data from airport, public, and their own weather stations around the region] is dramatic, massive,” Wadi says, “around half a terabyte per hour.” With JordanWeather.jo, weather information was collected and analyzed using a more local, on-the-ground strategy.
The team began renting processing power from supercomputers in Helsinki, Slovenia, and more recently, Belarus. The weather stations send raw data for the computers to put into models (the WRF model among others), and receive back images, numbers, and weather pattern simulation forecasts that the meteorological and weather teams then translate into content.
Two for the price of one
Utilizing Node.js, Wadi says, the team was able to get an entirely new product up and running in just two days. Their speed paid off: ArabiaWeather ended up selling the result, branded Aviator, which sends notifications specifically tailored to pilots and the information they need to safely fly (like wind speed and direction, visibility, etc.) to Royal Jordanian.
The team’s ability to use this technology all at once is contributing to deals like that with Royal Jordanian, as well as one with a “global player” in the oil and gas industry. “A main problem [these companies] face is windspeed and storms at sea, which threaten their offshore rigs,” says Wadi. “If a company has a rig out at sea, and wind speeds reach over 70 mph, the thing has to shut down.” To avoid losing money on closing the rig, as well as “risking people going out in bad weather,” oil and gas operators need to know a very specific set of weather data over a specific time period.
Data – as well as the ability to process and communicate it – “is giving us an advantage when it comes to B2B,” Wadi says.
Up next on ArabiaWeather’s technological radar? “We want to begin using supercomputers right here in the region,” says the product specialist. “Our offices are here in King Hussein Business Park, which has the power and internet capabilities to host such a setup,” he says gleefully.