Data can be considered a modern business’s most valuable asset, and yet it’s not treated as such. Consider Uber and Lyft. Without data from billions of historical rides, traffic patterns, and outside sources, there could be no estimated times of arrival, no shared rides, and no estimation of a ride’s cost.
Despite data’s intrinsic worth, it’s still not included on a business’s balance sheets. Though data is valuable, it’s hard to assign such value, so it remains an “off-balance sheet” item.
Two main consequences of traditional accounting’s inability to quantify the value of data on balance sheets are inefficiencies and a lack of understanding of how data ties to business outcomes.
The lack of ability to quantify the value of data means data is left undervalued, underutilized, and under-monetized. To avoid this trap, companies can start by gaining visibility into their data and enabling their people to use it to drive revenues and profits. CFOs must start thinking about data as a balance sheet item to which they can apply metrics.
Wall Street uses a financial ratio called return on assets (ROA), which helps assess how efficiently a company uses its assets to generate profits. A higher ROA means companies are more efficient at using their assets for productive business purposes.
ROA provides a good benchmark for members of the C-suite and forces them to think long and hard about how to convert assets into profits. However, since data does not appear on balance sheets, it’s difficult to calculate the returns derived from investing in data.
If ROA is not the proper metric for calculating returns on data, perhaps there is a better way for companies to start quantifying the returns generated from using data appropriately: return on data asset (RODA).
The basic formula for RODA would be income driven by data assets divided by the cost to create and maintain data assets. Data assets can be a dataset, a notebook, or even an Excel model. The purpose is to provide a clear framework in which data projects can be compared within and across companies.
This basic formula quantifies a company’s ability to monetize the data they have cost-effectively. The principles of the formula require chief data officers (CDOs) to optimize above and below the divisor line.
The basic formula for RODA would be income driven by data assets divided by the cost to create and maintain data assets.
On the numerator side, CDOs must focus on use cases where data can be leveraged to drive business value. On the denominator side, CDOs must find the most cost-effective way to create (or acquire), maintain, and secure data assets; increase value from existing data assets; and decommission unproductive data assets.
At the micro level, the RODA framework helps CDOs and lines of business prioritize projects. On the macro level, RODA helps CFOs allocate capital more efficiently and helps chief technology officers (CTOs) make the right technology decisions. Simply put, if RODA is higher than the cost of capital, then a project creates positive economic value. If it’s not, the project is not worth pursuing.
Uber likely didn’t have to undergo the RODA exercise since data was embedded directly into its product. But most companies struggle to find ways for data to power their business. The first step would be ruthless prioritization and approaching data with the discipline of a financial metric, just like any other asset.
In practice, the exact attribution of costs and revenues will be difficult. RODA provides a framework for decision-making. The first step CFOs and CTOs should take is to learn about recent innovations in open-source software (OSS). There are OSS projects such as Delta Lake, Hudi, and Iceberg that enable data to become democratized across an organization.
Democratization, in turn, enables data to become assets that can be used to drive revenue and profits. Combining the RODA framework and technology enables data to come “off the balance sheet’ and helps the organization realize its full value.
Junta Nakai is global industry leader for financial services at Databricks.