When speaking with clients and colleagues about the promise of advanced analytics for medical device companies, I often hear that there’s not enough good data for analytics, or that internal data is patchy, incomplete, poorly defined and spread across many systems. Those with pharmaceutical experience complain of the lack of market data, such as physician-level prescription data sets used by pharma companies, which would allow medtech companies to have visibility into their own market share, along with full product usage and potential of their customers.


Yet people often don’t realize that medtech has lots of interesting big and small data sets internally—in some areas a lot more than other industries can ever hope for. Most medtech companies have very detailed and customer-specific invoice data that goes back for many years. Invoice data not only tracks purchases to accounts, addresses and, in many cases, individual doctors or departments, but is also available with detailed SKUs and packaging information. That allows for very sophisticated research of the customers’ purchasing patterns, including looking at SKUs that are frequently bought together or around the same time, which enables all sorts of predictive models for upsell, cross-sell, churn and SKU-level forecasting.


Many medtech companies also have detailed inventory data, complete with orders, scheduled cases and other forward-looking measures. Similar to retail companies, which extensively automated their supply chain and inventory management between 2005 and 2010, medtech stands to save millions of dollars in reducing inventory in all points of its vast supply chains while ensuring that the necessary levels of each product are always available when a customer wants it. When retailers such as Ann Taylor invested in predictive analytics to help manage inventory, the investment typically paid for itself in a single year or even a quarter.


Moreover, there’s the main medtech data asset—the IoT data streams that companies can embed into the devices to capture data about their patients and the usage of the products they sell. As the devices and their accessories become smarter, medtech is in the unique position to incorporate within the product design the data sets that their devices produce and capture, and then use that data for both internal research, packaging and monetization.


In addition to the internal data, there’s a wealth of publicly available data sets that can be used in all types of predictive modelling. U.S. CensusMedicare claims and reimbursements, and other government data sets, PubMedweather, and lots of sentiment and other social data can be downloaded or scraped from the open web and utilized in models from customer “potentialization” and influence/referral networks to predictive forecasting and launch planning.


Finally, there are lots of third-party data providers that package and sell a variety of data sets. IQVIA (formerly Quintiles/IMS Health) sells market-potential data for procedures and customer demographics, along with AHA, SHA and LexisNexis. Flatiron, MedMining, Humedica and Practice Fusion package and sell EMR data sets; LabCorp and Quest Diagnostics have extensive anonymized sets of lab data; and Nielsen, Experian and ComScore can provide detailed consumer demographics, in some cases with already embedded segmentations.


In short, the data really is all around us, with more and more entities capturing, producing and enriching the data. The Chicago Analytics Group estimates that global data volume will reach 37.2 zettabytes by 2020 (or 37.2 billion terabytes—a huge number). In our data modelling efforts for various clients, we found that with a little forethought and creativity, we can almost always find, scrape and buy enough relevant data to create significant improvement in the “predictiveness” of the specific questions that we’re trying to solve.


My advice to the companies looking to improve their advanced analytics capability is three-fold:

  1. Companies should create better visibility and a more strategic approach to internal data. A strategic data asset management leader or team should maintain a catalog of all of the internal data that resides in different departments and cross-reference it with potential analytics use cases. Many of the use cases such as customer churn, inventory management or cross-selling have a high enough ROI to warrant an investment into cleansing and warehousing the data. The initial effort doesn’t have to be big: an annotated repository of data along with a spreadsheet cataloging data sets with their analytics use cases is a great starting point, which many companies are yet to make.
  2. Companies should invest in research and acquisition of the external data sets. The number of providers and data products grows exponentially as more and more companies learn to collect and monetize their data, and it’s both a challenge to keep up with the data and an opportunity to find better and more predictive data. Those data sets that can improve models with multimillion-dollar impact may be worth continuous purchasing and incorporation into the companies’ data bloodstreams.
  3. Data assets should be governed and managed just as carefully as any other high-value resource within the company. I have seen many times how clients bought or developed some extremely valuable data set only to forget it and never use it again—a major loss of value. Proper data quality, security and governance procedures should be established to ensure that critical operational and analytics data is clean, secure and trustworthy. Both internal and enriched external data sets should be regularly reviewed for their potential monetization value, and partnerships with third-party data consolidators or other potential data consumers can add a healthy additional revenue stream to medtech companies.

In sum, the notion of a lack of data in medtech is a long-outdated myth. There’s plenty of internal, public and third-party data that can be used for advanced and predictive analytics. My advice to medtech companies is to start managing their internal data and external data acquisitions more strategically to derive significant returns on the assets that they already have.