Medical Case Study
Client Data Profile
• Digitally profiling cancer cells responses against different drug regimes
• Highly informative, data driven assays to increase the speed and accuracy of cancer treatments
• Large range of clinical and non-clinical data from hospitals, Biobanks, public sources, research machines and their own prediction-based algorithms
• Unique image data that can be acquired by companies for state-of-the-art machine learning algorithms
Objectives
• Find a quantifiable methodology to value data as part of a forward business strategy
• Track the trajectory of data value, correlating this with ROI
• Discover unique data insights that can be used to enhance new business lines
• Data value with respect to data sources and specific use cases
• Creating a new business line from Imaging Data
Our approach
Day 0-4
Framing Activities
• DVP considers the data sources, hubs/repositories, use cases, image data and replacement value of data. There is also an extra consideration for the ROI of data
• Initial meetings with the CDO and CEO to understand data, business, and subscription models
• Developing an understanding of the range of data sources, the volume of data and how data driven products produce financial value
• Gathering information on maintenance, acquisition costs and storage costs since company launch, to aid with ROI and data value calculations
Day 5-10
Collateral Activities
• Discovery phase: understanding moving parts, looking into the products in more detail, taking snippets of the data and analysing their structure, a deeper dive into the data model and more
• Select mathematically driven valuation methodology based on the discovery phase combined with research impacts of industry specific use cases, and domain knowledge
• Understanding data quality and unique data decay behaviour for cancer data through thorough research and working directly with the client
Day 11-15
Valuation Activities
• Executing data valuation techniques, ensuring defensibility and research backed weightings. Working with the client to ensure data is correctly represented in the calculations alongside the mathematics behind weightings and methodologies
• Calculating the ROI of the data, breaking it down by business lines using decision-based methods
• Producing unique insights for the client, including the price of a cancer model using data and market information • Reports and fine tuning to finalise valuation
• Tuning involved calculating which data value methods will be weighted higher due to business importance
Methodologies Utilised
A mathematical hybrid of:
Consumption Method: This was a cost-inclusive, consumption-based method, which relies on data ownership, a repository or platform with consumable data and consumers to consume the data.
Cost Method: Used to calculate the replacement or sale value of the data, including images.
Data Decay Model: A data decay model was selected based on key drivers.
Why it was selected
The consumption method was selected because unique data from various data sources were being contained in hubs and data repositories, they were also enhancing the value of the data by improving and combining data from these sources. Scientists, the client themselves and affiliated companies would consume this data for different use cases, including machine learning and algorithm development. As per the prerequisites of this method, there were both consumable data and consumers. Cost methods were used as the data was unique, owned by the client and had a potential commercial use, including image data. DVP identified vsale or replacement value via unit costs. Decision-based methods were used to identify the best use of ROI per model for 650 cancer models. They planned to develop 5000 more and wanted to estimate the prospective ROI.
Insights & Recommendations
• Calculating the value of a cancer model using data, enabled the client to understand how to price cancer models
• Data ROI, and methods to increases the value of the data without inducing extra costs
• Recommendations on how to create further value via use within machine learning
• Data monetization strategies revealed such as data licensing
• Supported enhancements of data governance policies and best practice