IBM and the Weather Channel: COVID-19 Tracker and Q&A with Watson
In March 2020, IBM Watson Advertising and the Weather Company launched the COVID-19 hub featuring the Q&A with Watson on the Weather Channel’s digital properties. With over 400M+ MAUs, the Weather Company is part of 40M people's daily habits. Keeping people safe during natural disasters is their brand DNA. Leveraging their reach, brand trust, and location data brings users easy access to critical COVID-19 information– down to the hyper-localized level.
Platforms: Desktop, Web, Mobile, Tablet, App, CTV, OTT, IoT, Display
Features: Interactive, AI, Voice, Video
The mobile COVID-19 Q&A with Watson is prominently featured in the 30-second national broadcast spot outlining IBM’s efforts to address COVID-19.
In the early days of the outbreak, people wanted to know the number of cases near them as well as the rate of change in cases and trends over time. And our functionality reflects that. The product team made the trend tracker based on amazing Cognos work and refreshed the data every 15 minutes.
Implementing agile work methods, the hub MVP took 8 days while working in parallel to design and develop the chatbot to be ready for launch in 4.5 days. Brian’s team continued to refine the experience throughout the process, working alongside partners on the core product team and leaders across Watson Health, Research, and Marketing.
A User Journey
1. For this New York City-based hub, a user’s location is dynamically populated with information relative to their county level. Utilizing the chatbot, users engage with specific voice, text, or simple topic buttons. The below examples start with a user exploring Signs & Symptoms.
2. The user is immediately presented with the most recent information on Signs & Symptoms from trusted, official sources. Information is presented in rich, interactive text, image, and video formats.
3. Upon scrolling to the end of each information module, relevant topics are presented for users to further explore and engage.
4. Tapping Warning Signs expands relevant information.
5. When a user types, “How many cases in Peekskill”? Watson understands the intent of the question to be for Peekskill, New York, and prompts the user to view information for that specific location.
6. Location data is instantly implemented and the entire hub page changes to now display relevant data for the county of Westchester in Peekskill , NY.
Multiple levels of IBM governance and legal were employed throughout our process. Once our data team organized the various approved data sets, we built the COVID bot on the Watson Ads framework. It’s fast, flexible, and allows us to test, learn and iterate quickly.
Throughout the time we’ve been live, we’ve been able to look at the constantly streaming data and react. Revising design, content, and structure in addition to adjusting training and adding new intents based on trends in the marketplace.
How did we do that? We looked at trends and shifts in topic interest and compared them to actual trending headlines and developments out of the government and media.
In addition, we built a tool to allow us to track the content on the CDC and WHO sites and get notified via slack when a change is detected. Providing clarity on the difference, so that we can determine quickly and proactively if an update is needed within the corpus.
Webviews for faster implementation across platforms. Potential partnerships with companies that can further empower and help consumers at scale. Anthem, CVS and State Farm have all inquired about using our bot in their O&O properties to provide trusted answers to questions about primary care, treatment and insurance claims.
Constant optimization of data and user inputs drive relevance for users. View high level IBM case study here.
A special moment for us was when Ginni joined our last stand up, she was so thankful and inspiring. It means a lot to be recognized not only by the national media and the public, but having the CEO of IBM join our scrum to humbly offer thanks was yet another sign of her powerful leadership skills.
Finally, Brian would be remiss for not thanking his team for everything they’ve done to make this cognitive tool a reality. Design is the intent behind the outcome. And while the Core Watson Cognitive Advertising Design team is made up of Robert Redmond, January Holmes, Juliana Guimaraes, and Georgios Saliaris, it also includes a village of PMs, OMs, developers, engineers, data scientists, and so much more…. Sincere thanks go to:
Arjun Dharma - Data and real-time scrape engine
Brian Luksis - System automation and QA
Chethan Rajkumar - Backend dev
Donna Byron - NLP engineering and training
Geetha Vasudevan - Direction engineering co-lead
Joanne Santiago - Content QA and content attribution
Juliana Guimaraes - Design support, state health dept DB build
Jacky Chan - Frontend API integration dev
Paul Matchen - Frontend development lead
Monica Fogg - Corpus development and training
Priya Pradeep - Backend development and scrape engine support
Sandeep Jaikaria - Web/mobile web QA
Sahana Subbanna - Frontend development
Vinusha Venugopal - Web/mobile web QA
Jay Lee - Backend data and systems integration
