Comcast and Machine Learning: Can One of America鈥檚 Most Hated Companies Revive its Reputation?
Comcast was named the most hated company in America. Can it leverage machine learning to improve its product offer and customer service?
Customers have grown increasingly dissatisfied with the value delivered by their cable providers, leading to the 鈥渃ord cutting鈥� trend (in which customers cancel their cable subscriptions from providers such as Comcast in favor of streaming services such as Netflix and Hulu) that has arisen in the United States. In 2017, the year-over-year decline in pay-TV subscribers reached an all-time high of 3.4% [1]. Although all cable providers are grappling with the cord cutting trend, Comcast is facing an additional hurdle: notoriously poor customer service. In an industry where customer service is underwhelming, Comcast scores worse than the industry average. In fact, Comcast was named 鈥淎merica鈥檚 Most Hated Company鈥� in 2017 due to the combination of high prices and poor customer service [2].
Having grown both organically and through mergers and acquisitions (such as its 2010 acquisition of NBC Universal) Comcast is one of the largest media conglomerates in the world, with revenues of $84.5B and net income of $22.7B. Comcast offers a variety of services including traditional cable, internet, home security, and mobile. With nearly 30 million customers, Comcast collects an enormous amount of data [3]. This data has provided Comcast with the opportunity to leverage machine learning to address the key issues it is facing in two ways. First Comcast is evolving its product offer to combat streaming companies and second, Comcast is improving its customer service to increase loyalty and retention.
In order to compete with streaming services, Comcast is using machine learning to refine its traditional cable product to provide a more customized viewing experience similar to what customers with Netflix subscriptions are familiar with. Specifically, Comcast is observing past behaviors, preferences, and content purchases to provide personalized recommendations. Comcast is also predicting, in advance, which programs will be popular and displays them to their customers on its 鈥渟mart menu鈥� [4].
In addition to creating a personalized customer experience, Comcast is also leveraging machine learning to create new features to increase the value it offers to its customers, such as its Emmy-award winning voice controlled remote feature that generated over 3.4 billion voice commands in 2016 alone. Comcast recognized that creating the feature with a manual process would not be scalable, given the breadth of programming available across its hundreds of channels. Instead, Comcast built a program to find the relevant shows and films its customers are looking for based on the few key words that they speak into the remote [5]. Since rolling out this new feature, Comcast continues to improve it, such as making it bilingual [6].
With respect to customer service, Comcast is using machine learning to predict customer issues and resolve them. Instead of a customer having to run through a host of troubleshooting techniques with a customer service representative on the phone, Comcast built a system that continuously receives real-time data to diagnose the issue a customer is experiencing and suggest the likely solution to resolve it [7]. In addition to improving the service experience from the customer鈥檚 perspective, Comcast is also saving money by being able to predict, with over 90% accuracy, whether the issue requires a technician to be sent out to a customer鈥檚 home to fix the problem. Comcast has estimated that it will save 鈥渢ens of millions鈥� of dollars by avoiding sending technicians to their customers鈥� homes unnecessarily [8].
Going forward, Comcast has the opportunity to build on its machine learning momentum in order to continue improving both its product offering and customer service. As noted above, one of the key issues facing Comcast and other cable providers is the perceived lack of value that traditional cable companies provide relative to lower-cost streaming alternatives. Comcast should enhance its personalized recommendations to not only suggest popular, trending programming but to also nudge customers to watch shows and films on channels that they haven鈥檛 viewed programming on in the past, so that the customers can recognize the value in having access to a wide range of channels.
Additionally, because Comcast owns NBC Universal, Comcast should leverage the vast customer preference and behavior data it has across the hundreds of channels it provides, including premium channels such as HBO and Showtime, to feed into NBC Universal鈥檚 content creation process. If Comcast can learn to better predict what types of programming viewers will respond favorably to, it can provide an important input into the process of creating new hit shows and films.
Will providing a product similar to the streaming experience, or adding convenient features such as the voice-controlled remote, be enough to save Comcast from the 鈥渃ord cutting鈥� trend? Can machine learning help Comcast reinvent its brand or is one of 鈥淎merica鈥檚 Most Hated Companies鈥� doomed by its historically poor service?
(word count: 780)
References
[1]聽Pressman, A. (2018).聽http://fortune.com. Fortune. Available at: http://fortune.com/2018/03/01/cord-cutting-record-internet-tv/.
[2]聽Stebbins, Samuel, Evan Comen, Michael B. Sauter, and Charles Stockdale. 2018. “Bad Reputation: America鈥橲 Top 20 Most-Hated Companies”.聽Usatoday.Com. https://www.usatoday.com/story/money/business/2018/02/01/bad-reputation-americas-top-20-most-hated-companies/1058718001/.
[3]聽“2017 Comcast Annual Review”. 2018.聽Comcast Corporation. https://www.cmcsa.com/financials/annual-reports.
[4], [7] “Operationalizing Machine Learning At Comcast”. 2018.聽H2o.Ai. https://www.h2o.ai/wp-content/uploads/2017/03/Case-Studies_Comcast.pdf.
[5]聽“Comcast Wins Emmy Award For X1 Voice Remote Technology”. 2018.聽Corporate.Comcast.Com. https://corporate.comcast.com/news-information/news-feed/comcast-wins-emmy-award-for-x1-voice-remote-technology.
[6]聽Baumgartner, Jeff. 2018. “How Comcast Used AI To Make Its X1 Voice Remote Bilingual”.聽Light Reading. https://www.lightreading.com/artificial-intelligence-machine-learning/how-comcast-used-ai-to-make-its-x1-voice-remote-bilingual–/d/d-id/746670.
[8]聽Dano, Mike. 2018. “Comcast鈥檚 Machine Learning App Could Save 鈥楾ens Of Millions鈥� Of Dollars In Truck Rolls | Fiercevideo”.聽Fiercevideo.Com. https://www.fiercevideo.com/cable/comcast-s-machine-learning-app-saves-tens-millions-dollars-truck-rolls.
Photo: https://www.businessinsider.com/comcast-earnings-beat-sheds-140000-cable-tv-subscribers-2018-7
The two innovations you point out are quite creative but also novel. As the trend movers more and more to streaming and new value add content, Comcasts has a tough hill to climb. With all the data that they have, its interesting they haven’t tried to create algorithms that further personalize the customer experience and give individualized pricing.
I think Comcast’s attempt to apply machine learning to the problems of cord-cutting and poor customer service makes sense, but I wonder if this is a case of too little too late. Unless Comcast is able to provide consumers with customized offerings that are at a substantial discount to their current content offerings, I am not convinced that customers will be willing to continue paying their high prices. I do think applying machine learning to their customer service problem will enable Comcast to create value for consumers by anticipating customer needs and providing effective solutions in a more timely manner. However, I am not convinced that you can fully remove the human element from customer service. Comcast would do well to supplement its investments in machine learning with human capital investments to further improve the human interactions inherent to the customer service experience.