Feature
Machine learning (ML) allows companies to analyze vast troves of data to discover statistical patterns. Once trained, an ML model can take new data and predict how it fits those patterns and which outcomes can be expected. This can mean collecting data on sales, supply chains, prices and other factors and using the data to set prices, predict outcomes, develop products and design customer experiences. Here, we look at some examples of how ML is being employed to direct strategy and create results.
1. Mercy
Mercy, a health care system with more than 900 locations and more than 40,000 employees, wanted to modernize its data infrastructure to bring down costs and provide better outcomes for patients. To achieve this, the organization worked with Microsoft, which had an existing relationship with Epic, the company that handled Mercy’s electronic health records (EHRs), according to a case study.
Mercy used Azure Data Lake to bring all its data, which came to nearly 5 petabytes, together in the cloud. With their data no longer in silos, Mercy applied Azure Machine Learning to look at the information and predict next steps and outcomes for patients, providing more personalized care.
The algorithms help to predict which patients are at risk for particular conditions and suggest the best actions to take in caring for a particular person. Among other outcomes, it helps to determine when to send patients home. Automating tasks, such as appointment confirmation, frees up staff to focus more directly on patient care.
“We’re using … tools to do advanced data science around outcomes for conditions like hypertension and congestive heart failure,” says Brian Albrecht, VP of technology strategy, Mercy.
Results
- Reduced hospital stays by thousands of days over the course of a year
- Created smart dashboards to give providers more insight into a patient’s condition
- Created a model to identify information on insurance cards and automatically place it into EHRs
2. Randstad
Randstad, a Netherlands-based human resources services provider, operates in 39 countries. It wanted to upgrade its data analytics, including giving its sales team faster access to more complete and accurate data that was previously separated into silos, according to a case study.
The company decided to work with Google Cloud, which it had previous experience with, and ML6, a Google Cloud partner in Belgium that provides companies with machine learning services and engineering support for analytics.
ML6 helped Randstad use Google’s Big Query analytics engine to combine its data on existing and prospective customers with information from external sources, such as job postings. A machine learning model took that data and provided insights into the job market for employees. That gave the company a specialized customer relationship management (CRM) system, which it named Signal.
"We’re making great strides in more fully understanding our customers and offering our services at ideal times," says Mar Kessler, global VP of digital strategy and innovation, Randstad. "This translates to better relationships with existing customers and a stronger start with prospects."
Results
- Improved sales outreach success from 25% to 70%
- Reduced demands on IT management and maintenance by moving to a cloud system
- Scaled from a pilot program in two countries to providing ML access in 39 countries
3. READDI
After the COVID pandemic, READDI wanted to identify which virus family may be responsible for the next outbreak and develop antiviral drugs to be ready for it. The company began as the Rapidly Emerging Antiviral Drug Development Initiative, a collaboration among researchers at the University of North Carolina at Chapel Hill, Eshelman Institute for Innovation and the Structural Genomics Consortium.
After working with government, industry and philanthropic organizations to tackle COVID outbreaks, READDI turned its attention to future virus outbreaks, according to a case study.
The company worked with the SAS Analytics Center of Excellence to manage its large volume of data about various viruses, including coronavirus, alphavirus and flavivirus. It then tackled the data with SAS Viya, which uses machine learning to automatically create analytical models. READDI aims is to find two broad-spectrum antivirals per virus family.
The company also combined its data with data collected from patients with severe COVID and from EHRs and found how individual patients vary in response to the same virus, which could help it develop personalized therapeutics.
“It’s been exciting to see new things we never would have thought to look at become critical in some cases to our understanding of disease,” says Nat Moorman, co-founder and scientific advisor, READDI.
Results
- Identified new targets for antiviral drugs
- Gained new understanding of individual patient response
- Set on a path toward personalized viral therapeutics
4. IBM
IBM wants to set the best prices for its products, maximizing revenue while not driving away potential customers. It’s a complex task that requires many hours of manual assessment and a lengthy approval process, and it can be difficult to account for variations in geographic, business and market trend data, according to a case study.
IBM decided to use machine learning to solve these problems. The company created its Cognitive Pricing Analytics system to recommend optimal pricing for each individual customer based on their geographic location and buying patterns. It looks at years of data about past transactions and then uses machine learning models to weigh all the factors and predict the best price. It does this not only for IBM, but for its business partners as well.
The solution uses customers’ historical buying behavior and incorporates the latest market trends, such as product pricing strategy, currency conversion and inflation rate, to “recommend the price for winning a bid,” using regression-based machine learning models, says Nitesh Garg, CIO AND? data scientist, IBM.
Results
- 41% of generated quotes were accepted by clients
- Price quotes available within minutes
- Removed one level of required approval
5. Sky UK
Sky UK, which provides internet, cable television, telephone and mobile services, wanted to give its customers a more personalized experience. The company could divide up its 22.5 million users by favorite television genres but that would provide categories that are too broad to truly focus on their interests, according to a case study.
For customer personalization, Sky UK turned to machine learning tools from Adobe. Sky UK started by integrating all of its data about customer interactions — whether by phone, in person or online — using Adobe Experience Cloud. The company’s analytics team previously tried to sort customers into different segments by writing rules to assign them to one category or another, but even with Experience Cloud, that approach was time-consuming and did not always cover all possible combinations of interests.
Sky UK started using Adobe Sensei, a machine learning framework, to analyze customer data in real-time and provide recommendations for services. It also used Segment Compare, powered by Sensei, to analyze characteristics, such as which products they own, where they live and whether their previous interaction with the company was successful. The results allow the company to route the customer to the service representative who can best handle their issue.
“Brands that do not take advantage of machine learning capabilities risk providing disconnected experiences by failing to personalize interactions with their valued customers,” says Rob McLaughlin, head of digital decisioning and analytics, Sky UK.
Results
- Created hyper-focused segments of customers
- Delivered actionable intelligence and improved customer relationships
- Connected data from different channels to understand customer needs
