Today’s enterprises have an enormous amount of data with which to work, including traditional, structured data, as well as dark data that enterprises are now lighting up with breakthroughs in natural language processing, data extraction, and classification technologies.
Capitalizing on insights derived from data, users across an enterprise can make better decisions, evaluate risk, and find ways to engage and keep customers. While this information is certainly valuable, what is the exact value of enterprise data? Many CIOs and C-suite executives have been asking this question. In our experience working with Fortune 500 companies, we recommend a structured approach to put a value on enterprise data. It is built on three layers: intrinsic value, derivative value, and algorithmic value.
Often companies can aggregate their information and make it available in a data-as-a-service model. For instance, Equifax aggregates payment track records and sells credit history information to institutions looking to support financial decisions. Similarly, Dun & Bradstreet has a large database of commercial data on companies that it sells to businesses for evaluating suppliers and partners. Moreover, retailers that have extensive data on shoppers can sell the information to consumer packaged goods companies and other manufacturers. Some telecommunications companies sell GPS or consumer data through licensing in either aggregate or individual formats. The monetized value of this data using a discounted cash flow approach provides a measure for intrinsic value.
Derivative value is the notion of recombinant innovation, where an organization combines a set of data with other information to create new value. For instance, an agricultural company can combine weather data organized by zip code with data on the soil and weeds by zip code, as well as the seed attributes. The company can then distill this information to provide optimized fertilizer and pesticide recommendations that optimize crop production.
In another instance, banks can apply a basic graph algorithm (such as the one LinkedIn uses to recommend people to one another) to financial transactions in order to identify a network of money launderers, and then combine this data with regulatory data (such as beneficial ownership), sensor data (such as point of sale charges and phone calls), and web scraping data (such as web searches). Valuing this data requires business model thinking, evaluating adjacencies, and building a model based on domain derivative value.
Lighting up dark data and finding its derivative value is a great way to capitalize and take advantage of once intangible information. For instance, an enterprise can aggregate a batch of company emails and extract data, such as destination, sender, subject, and time stamps. Then, it may uncover a strong correlation between employee performance and certain email patterns. The company can analyze the relationship among these variable and do early identification of its high performers in its workforce.
One of the most exciting ways to leverage data is prediction. Netflix does an amazing job at this. It collects data on consumer behavior and applies machine learning to predict what someone would likely watch next. As people watch more, the recommendations get smarter. By matching content to viewers’ tastes, Netflix builds loyalty and drives topline growth.
Similarly, insurance companies can collect libraries of photographs from car accidents and claims information for each case, apply them as labeled data to a machine learning engine, and predict claims payouts. Whereas before customers would have to wait for a lengthy assessment, visit from a representative, etc., now they can quickly find out what their claim payout will be and determine the appropriate next steps, thereby simplifying their overall experience with the insurance company. In another example, a commercial lender can use data from previous loan transactions to develop a recommender model to determine which financial product in which configuration the lender’s sales agents should suggest to a specific borrower.
In all these examples, the true value of data comes out when enterprises apply generic algorithms (such as graph traversal algorithms, segmentation algorithms, or entity extraction algorithms) to a specific domain data set. This bridges the last mile between algorithm theory and industry domain.
In summary, the value of data is getting increasing attention from boards and CXOs as digital technologies disrupt entire industries. Getting a handle on the enterprise value of data is becoming more important for most corporations. A structured approach exists today for valuing data and it involves aggregating three components of value – intrinsic, derivative, and algorithmic.