In science, when more than 50 percent of a substance has undergone a radical change, it’s called a half-life. The term is commonly used in measuring radioactive material, but the same concept is increasingly being applied to describing the changes in data and its usefulness.
A recent publication by Nucleus Research outlines the philosophy that data has an initial value that decreases over time. The brief, entitled “The Half Life of Data,” suggests that companies use the customer-centric approach for data distribution. More importantly, researchers heavily endorse analytics tools to extend the life of data and improve the effectiveness of decision making.
“We used the scientific concept of half-life to model this data decay based on three different decision-making cadences: tactical, operational, and strategic,” says Anne Moxie, a senior analyst with Nucleus Research. “End users for these different decisions range across the full spectrum, and we wanted to evaluate how each of them is best provided with data accessibility.”
It makes sense if you consider your own use of data. Sometimes you only need to remember facts at your fingertips while other times require long-term memory and analysis. Consider the picture below. The graph represents a network of 2,724 Twitter users whose tweets in the requested range contained "#oracle", or who were replied to or mentioned in those tweets. How long would you consider holding onto that data?
Data: From Fleeting to Long-Lived
If data is being used for tactical decisions—ones that are read off of dashboards—it’s likely that business decision makers use that data to make incremental choices and that the information is very fleeting. These data points are often generated by enterprise applications affecting customer, employee, or supply chain management. As represented in the graph below, Nucleus estimates the half-life of tactical data occurs after a mere 30 minutes.
Not as temporary, yet nearly as time sensitive, operational data tends to gets stale after about eight hours, according to Nucleus. This type of data is important for weekly decisions. Say, for example, a grocery store manager wants to explore the popularity of a product and budgets appropriately so his shelves are stocked in a handful of different areas around the store. The data necessary for implementing this action requires some flexibility and implies the need to create data visualizations to prove its worth.
The longest-lasting type of data is called strategic by Nucleus. Data falling into this category includes information that is used to forecast and project business opportunities. Nucleus researchers recommend employing specialists to improve data preparation, as this type of data requires better capture and distribution.
“If software solutions are not being used, then customers will not see a positive ROI,” says Moxie. “This has been a particular challenge for analytics deployments because there are many hands that are involved in analytics use, ranging from business users to data scientists. Customers need to make sure that their deployments offer accessibility to data, and therefore, they should have a user-centric focus on deployments instead of a data-centric focus.”
Although Oracle has the analytics tools capable of addressing the half-life of data, it bears repeating that organizations must start the process with a hard look at who is using the data to make decisions and how long that data be relevant. Otherwise, as Nucleus Research explains, they risk “a poor ROI from their analytics deployments.”
Featured image courtesy of Nucleus Research.
Embedded image courtesy of NodeXL Graph Gallery