\u201cKnowing how weather impacts consumers\u2026it has an influence on what products you\u2019re going to buy, what activities your going to engage in, [and] your mood \u2013 that\u2019s at the heart of what we do at AccuWeather\u201d [1] \u2013 Rosemary Radich, Director of Data Science at AccuWeather.<\/p>\n
In the last few years machine learning has been transforming the capabilities of the weather forecasting industry. For example, in 2016, IBM purchased the Weather Company and introduced the capabilities of Watson into the industry [2]. Responding to this competition, in 2017, AccuWeather [3] launched D3<\/sup> (Data Driven Decisions), a cloud-based machine learning platform powered by Microsoft Azure that enables businesses to predict how weather will impact important aspects of their business, including sales and operations [4]. Originally entering the space around 2012, this signaled a major leap forward for the company [5].<\/p>\n In this easy-to-use D3<\/sup> platform, a business can upload historical sales data for specific products and within minutes use a dashboard to see how sales for those products are affected by weather such as wind, snow, rain, or humidity [6]. \u00a0While AccuWeather had previously analyzed how the weather impacted their client\u2019s business performance such analyses could take months to perform [7]. Radich notes that machine learning enhanced this type of analysis making possible the creation of \u201cnew products and services to improve our offerings\u201d [8]. Adopting machine learning was critical for AccuWeather to remain competitive in the B2B weather forecasting space.<\/p>\n The applications are limitless. D3<\/sup> Express uses ten years of weather data to predict how disruptive a weather event will be on a business using a scale of 1 (insignificant) to 10 (extreme) [9]. When employed, a supply chain manager can be notified of a pending weather event\u2019s impacts and alter shipping routes to ensure on-time deliveries [10]. Marketing teams can implement promotional events to drive the sale of items that are positively correlated to the weather [11]. Operations teams can match staffing and inventory levels to meet shifting demand [12]. The end result is increased revenues and reduced costs.<\/p>\n AccuWeather\u2019s push into predictive analytics and machine learning is already yielding results. For example, they worked with Starbucks to help solve the seasonal problem of running out of ice and cups in hot weather [13]. They worked with a global candy manufacturer to identify if their sales spikes were weather-related [14]. AccuWeather partnered with Spotify to analyze 85 billion music streams to learn how different weather in various cities impacts the listening habits of their customers [15]. The end result was Climatune, a new music platform that features songs based on the weather-induced mood for the day [16].<\/p>\n Going forward, AccuWeather is focused on rapidly moving all of its weather and non-weather data onto the Microsoft cloud so they can continue to explore the relationships between weather and consumers [17]. They are also building \u201cbots\u201d that combine both machine learning and language processing to provide weather insights in a human fashion [18]. \u00a0For example, AccuWeather recently launched an AI Weather Bot for Facebook Messenger [19]. They are also looking to use machine learning in imagery analysis of satellite and radar to increase the accuracy of historical, current, and forecasted weather data [20].<\/p>\n Internally, they are using machine learning insights to improve their own business [21]. For example, the insights from machine learning could be used to send targeted, weather-appropriate advertisements to their roughly 1.5B worldwide app users, increasing revenues for their advertising customers and the rates they can charge them [22].<\/p>\n To remain competitive AccuWeather should leverage its strength: its years of experience in forecasting and climatology. Machine learning algorithms are only as accurate as the data they are \u201cfed\u201d. Cleaning meteorological data is increasingly challenging as the data becomes more granular but also greater in scope [23]. Deep industry knowledge is critical in this data scrubbing process [24]. Other companies have machine learning capabilities but few, if any, have AccuWeather\u2019s experience with weather data. This experience should allow them to create better data inputs and ultimately better predictions for their customers.<\/p>\n AccuWeather must lead in using machine learning to improve critical weather forecasts. For example, machine learning can compare model [25] forecast outputs to historical weather data to determine situations when a computer model is more or less likely to be correct. NOAA scientists found that applying AI techniques to models coupled with sound meteorological practice can improve prediction skill for high impact weather [26]. If AccuWeather\u2019s forecast accuracy increases and extends further into the future (since increasing notice by one or two days can make a big difference in some industries), they will be differentiated in the market [27].<\/p>\n While the development of D3<\/sup> has enabled AccuWeather to provide new products and insights for their customers, what other markets should AccuWeather pursue with this technology? Will there be more opportunities to expand in consumer social media and travel apps, or should they focus on B2B clients in industries such as retail, agriculture, and insurance?<\/p>\n (797 Words)<\/p>\n [1] Microsoft Cloud. (2018). \u201cPredicting weather impact using cloud-based machine learning.\u201d https:\/\/www.youtube.com\/watch?v=ImwXpbI5Ino<\/a>, Viewed on 10 Nov. 2018.<\/p>\n [2] John Walker. (2017). \u00a0\u201cAI for Weather Forecasting \u2013 In Retail, Agriculture, Disaster Prediction, and \u00a0\u00a0More.\u201d TechEmergence<\/em>. https:\/\/www.techemergence.com\/ai-for-weather-forecasting\/<\/a>, Accessed 10 Nov. 2018.<\/p>\n [3] AccuWeather was founded in 1962 and delivers weather forecasts to nearly 2 billion people worldwide and to about half of Fortune 500 companies. Accuweather. (2018). https:\/\/corporate.accuweather.com\/about<\/a>, Accessed 10 Nov. 2018.<\/p>\n [4] AccuWeather Press Release. (2017). \u201cAccuWeather Launches AccuWeather D3 Express at Microsoft Build 2017.\u201d https:\/\/www.accuweather.com\/en\/press\/65947535<\/a>, Accessed 10 Nov. 2018.<\/p>\n [5] Microsoft Customer Stories. (2016). \u201cWeathering the storm: Big data and cloud technologies safeguard lives and businesses worldwide.\u201d https:\/\/customers.microsoft.com\/en-US\/story\/accuweather<\/a>, Accessed 10 Nov. 2018.<\/p>\n [6] Microsoft Customer Stories. (2018).\u00a0\u201cBusinesses predict weather impact using cloud-based machine learning\u201d. https:\/\/customers.microsoft.com\/en-au\/story\/accuweather-partner-professional-services-azure<\/a>, \u00a0Accessed 10 Nov. 2018.<\/p>\n [7] Microsoft Cloud. (2018). \u201cPredicting weather impact using cloud-based machine learning.\u201d https:\/\/www.youtube.com\/watch?v=ImwXpbI5Ino<\/a>, Viewed on 10 Nov. 2018.<\/p>\n [8] Microsoft Customer Stories. (2018).\u00a0\u201cBusinesses predict weather impact using cloud-based machine learning\u201d. https:\/\/customers.microsoft.com\/en-au\/story\/accuweather-partner-professional-services-azure<\/a>, \u00a0Accessed 10 Nov. 2018.<\/p>\n [9] AccuWeather Press Release. (2017). \u201cAccuWeather Launches AccuWeather D3 Express at Microsoft Build 2017.\u201d https:\/\/www.accuweather.com\/en\/press\/65947535<\/a>, Accessed 10 Nov. 2018.<\/p>\n [10] Ibid.<\/p>\n [11] Ibid.<\/p>\n [12] Ibid.<\/p>\n [13] Microsoft Customer Stories. (2016). \u201cWeathering the storm: Big data and cloud technologies safeguard lives and businesses worldwide.\u201d https:\/\/customers.microsoft.com\/en-US\/story\/accuweather<\/a>, Accessed 10 Nov. 2018.<\/p>\n [14] Ibid.<\/p>\n [15] Zach Brook. (2017). \u201cMarketers Are Using the Weather to Predict Buyer Behavior.\u201d American Marketing Association<\/em>. https:\/\/www.ama.org\/publications\/MarketingNews\/Pages\/how-marketers-are-using-the-weather-to-predict-buyer-behavior.aspx<\/a>. Accessed on 10 Nov. 2018.<\/p>\n [16] Ibid.<\/p>\n [17] Microsoft Customer Stories. (2018).\u00a0\u201cBusinesses predict weather impact using cloud-based machine learning\u201d. https:\/\/customers.microsoft.com\/en-au\/story\/accuweather-partner-professional-services-azure<\/a>, \u00a0Accessed 10 Nov. 2018.<\/p>\n [18] Microsoft Mechanics. (2017). \u201cHow we built-it: AccuWeather on architecting their self service weather delivery platform on Azure.\u201d https:\/\/www.youtube.com\/watch?v=dYjWDOzag_U<\/a>, \u00a0Viewed on 10 Nov. 2018.<\/p>\n [19] Shimonti Paul. (2017). \u201cArtificial Intelligence Powered Weather Bot for Facebook Messenger.\u201d Geospatial Media and Communications<\/em>. https:\/\/www.geospatialworld.net\/blogs\/artificial-intelligence-powered-weather-bot-facebook-messenger\/<\/a>, Accessed on 10 Nov. 2018.<\/p>\n [20] Microsoft Mechanics. (2017). \u201cHow we built-it: AccuWeather on architecting their self service weather delivery platform on Azure.\u201d https:\/\/www.youtube.com\/watch?v=dYjWDOzag_U<\/a>, \u00a0Viewed on 10 Nov. 2018.<\/p>\n [21] Ibid.<\/p>\n [22] Zach Brook. (2017). \u201cMarketers Are Using the Weather to Predict Buyer Behavior.\u201d American Marketing Association<\/em>. https:\/\/www.ama.org\/publications\/MarketingNews\/Pages\/how-marketers-are-using-the-weather-to-predict-buyer-behavior.aspx<\/a>. Accessed on 10 Nov. 2018.<\/p>\n [23] Editorial Team. (2017). \u201cThe New Age of Analytics: Artificial Intelligence and Data are Note Enough to Power Your Business.\u201d Inside BIGDATA<\/em>. https:\/\/insidebigdata.com\/2017\/05\/23\/new-age-analytics-artificial-intelligence-data-not-enough-power-business\/<\/a>, Accessed on 10 Nov. 2018.<\/p>\n [24] Ibid.<\/p>\n [25] Common forecasting models used in weather prediction include the Global Forecast System (GFS) model and the European Center for Medium Range Weather Forecasting Integrated Forecast System (ECMWF) model.<\/p>\n [26] Amy McGovern, Kimberly L. Elmore et al. (2007). \u201cUsing Artificial Intelligence to Improve Real-Time Descion-Making for High-Impact Weather. American Meteorological Society. \u00a0https:\/\/journals.ametsoc.org\/doi\/10.1175\/BAMS-D-16-0123.1<\/a>, Accessed 10 Nov. 2018.<\/p>\n