The Definitive Guide to Data Recency

data recencyAs data plays an expanding role in enterprise sales and marketing, discussions around data recency have increased. As a leading B2B data provider for sales and marketing organizations, here at ZoomInfo we’ve had the opportunity to study data recency firsthand.

Let’s start with this: The concept of data recency is fluid and difficult to measure. Why? Because in dynamic datasets, information is updated and appended on a constant basis. At what point can one say the dataset has been refreshed? Is it when all records have been updated in some way? If so, is the refresh carried out once or on an ongoing basis?  Is there a percentage that qualifies as a refresh? Do all fields associated with an individual need to be updated or does updating a single field count?

As you can imagine, at a certain point you end up in an angels-on-the-head-of-a-pin scenario. Rather than viewing recency as a finite thing that can be achieved at measurable intervals, let’s rethink data recency altogether. What matters most is whether the most critical data is correct where and when you need it.

Applications of Varying Data Recency

Let’s look at four scenarios that describe different rates of recency:

Spotify: Spotify currently has about 35,000,000 songs and adds about 20,000 every day. For every song, there are a significant number of data points. These include artist, personnel, producer, title, length, genre, copyright, etc. Many of them are static. As a fan, I don’t care if they haven’t re-updated track length for an album that’s 20 years old, I want to see the newest tracks as soon as they’re available.

Fantasy sports: Fantasy sports live and die by data. In some cases, information needs to be up-to-the-instant fresh but, in other situations and use cases, stats are pulled from the past and mixed and matched to create new numbers, valuations, and decisions. Once my team is submitted, I only really care about accurate updates on my fantasy team and the team I’m playing against. If I’ve clinched my contest on Sunday, I don’t care about data from the Monday night game.

Quarterly financial data: Quarterly financial data is open to interpretation. Different stakeholders (corporate executives, board members, investors, feds, bankers, business reporters) will each interpret the data in a different way. The numbers may be fixed in that snapshot, but the buy/sell/hold decisions will vary based on multiple factors—many beyond the market numbers.

Algorithmic trading: In an example from the world of fintech, algorithmic trading is the ultimate test of recency since even a few fractions of a second can mean big gains or losses. This type of data has virtually no value outside of a singular moment. (It can, of course, be used for testing and evaluating historical performance.) Recency, in this case, equals immediacy.

In each of these scenarios, there is a different definition and expectation of recency and an implicit acceptance (or non-acceptance) of some degree of latency. Recency needs to be considered through the lens of usage. To apply a fixed definition of an attribute of data ignores the reality of different use cases.

Rolling Recency

For many applications—especially in the world of large and dynamic datasets – the idea of rolling recency—is worth consideration. This approach seeks to update critical data on a constant and ongoing basis. The strength of this model is that data is under constant scrutiny and is updated as soon as new information becomes available. New information, however, may not be created on a regular basis.

The irregularity of anticipated change could be a drawback of the rolling recency model. There are some who want a date-stamped version of their information. Being certain that at X time on Y day the data is fresh and locked down as the “most recent” can be a comfort. That is true so long as one is comfortable with static data between update points.

It is critical that organizations understand and have faith in the data that fuels their businesses. That faith needs to be based on context. There are degrees of data recency, with some forms being more important than others. As businesses consider their own data needs, they need to decide which data is most important and how current that data must be. There are many ways to measure the value of recency and a single standard simply may not be a fit for everyone.

For instance, ZoomInfo is the only platform reading and parsing data at unparalleled speed and scale (think a billion-plus data points every month). This is accomplished through a combination of machine-learning algorithms, an automated system of checks and balances, and third-party verification tools to provide up-to-date information on a continuous basis.

This is a more effective approach than manual processes to gather, update and verify data. And, as machine learning matures, this approach will be the way in which the industry moves. To use an analogy, while some companies update their data periodically, akin to the old phone book, we are constantly updating information and pushing it live for our customers.

Key Considerations for Better Data

In comparison to recency, the areas of data availability, and the quality and quantity of available datasets are considerably easier to define and measure. The importance of having access to an extensive database of accurate data is obvious: The more detailed, accurate, and timely your data is, the more effectively you’ll be able to put that information to work.

Those tasked with assessing and validating data must first understand how that data will be used. Does it need to be current on a second by second basis? Are there some classes of data that can be refreshed on a longer-term basis? Answering these questions requires bringing together all data stakeholders in your evaluation for frank and open conversations on how data is used within an organization.

The trio of quality, quantity, and recency are the core metrics in data intelligence. Organizations that are serious about their use of data must use them to assess the strength of their own information assets. These three elements determine the value of a dataset and ultimately provide a lens for understanding the successes or failures of data-fueled initiatives.

Contact our sales team today to learn more about our extensive company and contact database. We have the solutions you need to scale your business growth.

About the Author: Nir Keren, Vice President of Research & Development, is the driving force behind ZoomInfo’s technological innovations. An entrepreneur at heart, Keren was the Founder and CTO of AdSAP, an all-in-one online advertising platform.

He was also the founder and CTO of ONDiGO, a sales-automation platform, which automates tedious sales tasks and provides salespeople with a daily personalized action list for closing more deals. Prior to that, he served as an embedded software engineer at Ceragon Networks in Tel Aviv.