Accelerating digital transformation

Data Science during Difficult Times

Across industries, the impacts of the COVID-19 pandemic compelled leaders to accelerate their company’s digital transformation and become more data-driven. For one global mid-sized pharmaceutical company that develops medicines tailored to address rare diseases, the journey began with a need to keep clinical trials on track amid travel restrictions, stay-at-home orders, supply chain disruptions, and widespread fears.

It led to the creation of a COVID-19 prediction dashboard, developed by management consultancy Arthur D. Little and built on the Domino® Enterprise MLOps platform. The dashboard enables operations staff to better select sites for trials in a rapidly shifting landscape and apply interventions. Since implementing the dashboard, company leaders sought to expand its insights—both to increase efficiency and effectiveness of clinical trials overall and to begin creating an "algorithmic backbone" that embeds data-driven capabilities throughout the organization for improved decision making.

"Hundreds of decision makers across clinical operations use the model to understand the risks related to the COVID-19 pandemic in each region and the potential impact on patient recruitment, site selection, and operations," said Michael Eiden, Global Head of AI and Machine Learning at Arthur D. Little. "It's been highly effective, and we've since begun building a network of machine learning models that tap into the different sides of the data and can consume each other's results as needed."

For Arthur D. Little, the Domino platform enables its AI team to develop these types of sophisticated machine learning and AI models for its clients more rapidly. "Good consultancy is about bringing diverse expertise to achieve a breakthrough outcome," explained Greg Smith, Partner with Arthur D. Little. "We're investing in our AI practice to leverage vast amounts of data to provide our clients with the best advice and effectively create new capabilities for them."

"The Domino platform is an enabler that accelerates our work. Without Domino, life is a lot harder, and some things may not be possible. It's a necessary component of our evolution and facilitates execution on our vision to deliver breakthrough outcomes for our clients."

—Greg Smith, Partner, Arthur D. Little

Challenge

As the COVID-19 pandemic began spreading across regions, executives at this pharmaceutical company met to discuss the potential impact on clinical trials. Clinical trials are a crucial part of the drug development process. They're also quite complex to manage with many moving parts, including patient recruitment, site selection, and supply chain management. Operational issues can dramatically increase costs, which often can put huge pressure on clinical teams as they try and manage patient care while trying to balance budget overrun, according to Arthur D. Little.

The company's existing forecasting models, which analyzed historical data from past trials, couldn't effectively account for the uncertainties in the marketplace. Executives had seen the cost of such gaps the previous year when market changes that went undetected quadrupled the overall budget for one clinical trial. Now, as the crisis loomed large, executives agreed they needed a more sophisticated forecasting capability—and they needed it quickly—if they were to minimize disruptions to clinical trials. 

Solution

"Our data scientists feel right at home on the Domino platform and appreciate how much freedom they have in experimenting in Domino."

—Michael Eiden, Global Head of AI and Machine Learning at Arthur D. Little

Company leaders engaged Arthur D. Little, which brought together an interdisciplinary team of data, engineering, and medical experts to solve the challenge. As part of their work, they developed and optimized a COVID-19 prediction dashboard that analyzes more than 100 data sources, including government, interaction, mobility, and demographic data. During the year, the Arthur D. Little team continuously refined the model, incorporating new data on vaccinations, virus strains, regulatory changes, and more to increase its predictive accuracy and support decision making in other areas of the company. For example, C-suite executives now have access to new decision support tools built on this network that enables them to evaluate the impact of potential regulatory changes across the company's entire research and development portfolio. 

COVID-19 PREDICTION DASHBOARD

Source: Arthur D. Little

Before launching this work, Arthur D. Little had begun work internally to implement a data science platform that would enable its team to collaborate and iterate more efficiently. They evaluated several solutions, ultimately selecting the Domino platform for its extensive capabilities designed for code-first data scientists.

"We wanted a data science platform for data science professionals who are at the leading edge of their discipline. Other platforms were too simple for hardcore data scientists. Our data scientists feel right at home on the Domino platform and appreciate how much freedom they have in experimenting in Domino. They are not constrained by CPU, hardware limitations, or tooling. And the more flexibility they have, the more creative they can be."

—Michael Eiden, Global Head of AI and Machine Learning at Arthur D. Little

Impact

Arthur D. Little's new COVID-19 prediction dashboard enabled the pharmaceutical company to understand how the pandemic's spread would impact clinical trials in different geographies, more effectively select sites, and improve budget management. "Previously, decisions were often based on best-case scenarios," said Raf Postepski, Principal with Arthur D. Little. "It led to significant deltas around budget, upwards at times in the region of 30 percent. They're now able to spot the signals in the data early and adjust their plans accordingly and keep clinical trials to plan."

While the pharmaceutical company's initial focus was getting a handle on the impacts of the pandemic on its clinical trials, company executives have widened their aperture to expand the use of machine learning across the business.

"The company saw how quickly we were able to turn initial ideas and concepts into working models, which has given them the confidence to increase their data science investment."

—Raf Postepski, Principal, Arthur D. Little

The Domino Effect 

For Arthur D. Little's AI team, the Domino platform is an important tool for accelerating problem-solving, in some cases developing new models up to 1.5 times faster.

"Domino helps us accelerate the ideation process significantly to bring new solutions to companies faster. If we weren't on Domino and still doing the things as we were previously, we'd probably be looking at 50 percent more time for some projects."

—Raf Postepski, Principal, Arthur D. Little

Increased team productivity has been a significant driver for decreasing project life cycles, driven by several factors, including:

• Reduced administration time and increased management efficiency with faster access to resources, automatic version controls, and standardized processes. "Before, we sometimes ran into hardware constraints, and we couldn't easily monitor how the models were doing during development because that plumbing didn't exist," said Eiden. "Domino has simplified all of this. I estimate data scientists are reclaiming between one-quarter and one-third of their time."
  
  
• More seamless collaboration among staff and with clients. Internally, data scientists can more quickly and easily exchange ideas and concepts. "We fundamentally think of data science as a team sport, and Domino enables us to create a frictionless environment for collaboration," said Smith. Externally, project teams can share models with clients more easily via APIs to obtain early feedback on model outputs and better shape the final product.
  
  
• Expedited onboarding of new team members as Arthur D. Little builds out a multi-location global Center of Excellence for AI and machine learning. "We've scaled our team quite dramatically in the last year, and with Domino, we've been able to get new team members quickly familiarized with the subject matter, code, and previous runs that people have executed,” said Eiden. “That has been a huge asset to us."
  
  
• Improved reuse with the ability to create building blocks of code that can be leveraged in different contexts to continually build upon past work. "Using Domino, we've been able to leverage ideas from one workstream in another across different industries," said Postepski. "That can help us do the next project even better and even smarter."