Understanding the Monetary Value of Data

 One of the biggest challenges that analytics departments deal with (beyond the technical challenges of extracting insights from data) is measuring Return on Investment (ROI) for analytics initiatives. While it may be relatively easy to understand the cost of analytics initiatives in terms of the cost of data storage and processing (e.g. cloud applications such as AWS, Microsoft Azure, and GCP as well as a myriad of BI tools in the market to choose from) it is much more complex to attribute improvements in business performance to analytics initiatives or to even start understanding the monetary value of the data companies possess. The book Infonomics by Doug Laney suggested a few interesting methods to understand the monetary value of data such as the Market Value of Information (MVI) or in other words, how many other companies in the market have the same data?

Understanding how to measure the monetary value of data becomes a hot topic for many reasons in today’s world, one of them is the application of digital product management in analytics that gets easier once such measurements enable companies to understand how to leverage their data internally and externally beyond plain business reporting and how to better develop analytics products. Here are some ideas for ways companies can start to understand the true monetary value of their data and move from seeing it as a cost that’s purely used for reporting to seeing it more as an asset, and I’ll use the case of the COVID-19 in order to illustrate some of this ways of determining monetary value (at the time of writing this blog post, COVID-19 has already became a pandemic):

1. Anonymous vs attributable data - Due to geographical data privacy initiatives such as the GDPR in the EU, and the CCPA in California, it is hard to collect data that is also identifiable without putting in place an authentication system. While anonymous data can serve internal reporting needs, its value is much lower than the one of attributable data since it is not actionable. Think about it this way, in the case of an e-commerce website you would be able to target specific customers by identifying those who logged-in to your website based on their online behavior and content preferences and you won’t be able to do the same thing with anonymous data. In the case of diseases, knowing how many people contracted a disease would allow you to report on it but knowing the exact location of clusters of infected people would allow you to take action in the form of scientific modelling that can help predicting the contagion speed of the disease and help inform authorities about the next steps to take.
2. Data quality – It’s a well-known fact that datasets that have quality issues lose their value. In the case of marketing operations, incomplete or corrupted data could cause delays in estimating marketing campaign operations. A good example could be missing or incorrect email addresses that can skew the results of a company’s forecasts regarding open and click-through-rate (CTR) of email campaigns. In the case of a pandemic such as the COVID-19, a large portion of missing data due to untested people who carry the disease can challenge optimizing the way of delaying the spread (“flattening the curve”) and skew the results reported in the media regarding the situation of the spread.
3. Means of extracting insights – The monetary value of data is also a relative measure and it varies between one organization to another. Some organizations can accumulate a lot of data overtime that could potentially be very worthy but lack the technological and human capabilities to extract any meaning from it. Data resembles oil in that sense, and that means that raw data is worth a lot less than processed and ready to use data. Think about organization A who is able to automatically capture, store, and curate large volume of data from social media about people impacted by disasters, extract it using simple SQL syntax, and transform it into insights to communicate to upper management right away using machine learning techniques such as sentiment and text analysis. In this case, the data organization A has is worth a lot, and not just in theory. In another scenario we look at organization B who can only reach a basic level of understanding about the big picture by manually extracting data from the same sources, but cannot generate insights as fast as organization A. In that case, the monetary value of the data captured by organization B is a lot lower even in the eyes of organization B itself since it has no capabilities in place to derive anything meaningful and actionable from it.
4. Time range of data – I believe that every industry has an optimal time range for a dataset that allows it to see seasonal and periodical trends (e.g. year-over-year, quarterly performance, etc.) and datasets that correspond to that length become more worthy. For a commercial business I believe it would be anywhere between 2-5 years since having less than two years of data makes it impossible to see seasonal and periodical trends in sales and customer interest. However, if a company generates a large volume of data it will be costly and inefficient for it to keep storing and processing this data since consumer trends could change drastically over this time. In the case of COVID-19 (or any epidemic or pandemic) we’re still in a place where we cannot see these trends due to a dataset that only contains numbers for a few months and because of that it’s hard for us to understand how the trend will continue. What we could do is to try to infer from past trends of similar diseases such as the SARS what the continuation of the spread will look like. Therefore, in the healthcare industry, longer time ranges of datasets can be valuable in order to try to make long-run predictions about the spread of diseases as well as their prevention and treatment.
5. Demand from the market – finally, any organization can generate a lot of data that can yield many interesting insights but when it comes to the real value of the data, the question that should be asked is how much interest and demand from the market exist for this data. A good example we can see in the digital marketing world is in the form of competitive intelligence tools for SEO keyword researchers such as SpyFu, Semrush, and Wordtracker, all offering insights about keywords competition, volume, and complexity, since the entire foundation of digital marketing (organic content and paid advertisement) is based on the selection of the right keywords so people and companies are willing to pay for this information. During the outbreak of a pandemic or an epidemic, any data related to the spread of past similar diseases become a commodity due to the lack of options to extract a lot of insights in the initial spread of the disease.

Understanding these concepts in relation to your specific industry can help you to start putting a monetary value of the data you own and see it more as an asset to your business rather than a cost.