Analytics isn’t an experimental application anymore. It’s time for IT to create maintenance and health practices to ensure long-term analytics quality.
Analytics, whether performed on structured or unstructured data, has reached a point of maturity in organizations to where it is being used regularly and, in a number of cases, as a mission-critical function. As daily use of analytics increases, there is also a tendency for the data and the algorithms used in analytics to get outdated, for security lapses to occur and for many organizations to lapse on aggressively monitoring if they’re really getting the value out of their analytics that they thought they thought they would.
SEE: Electronic Data Disposal Policy (TechRepublic Premium)
Situations like these have a number of implications for IT. The major implication is that IT must take analytics applications under its maintenance wing in the same way that it performs maintenance and safeguards the health of its transactional systems to meet the needs of the enterprise.
Here are four key “health check” and maintenance areas that IT should actively adopt to maintain the health of corporate analytics applications and data:
1. Have robust security
Is the security of your analytics applications meeting corporate standards? This is a critical question for many organizations running Internet of Things applications that stream data in real time. Security is a concern because a majority of IoT devices come into organizations with security presets that are significantly more relaxed than enterprises demand. The end result is that IT must “hand set” all of these IoT devices to the levels of security that the enterprise requires.
The risk is that it’s often easier for IT (or even end users) to just plug in new IoT devices and appliances and forgo extra steps like checking (and if necessary, recalibrating) the security on the devices. From January to June of 2021, some 1.51 billion breaches of IoT devices occurred, according to research conducted by Kaspersky.
2. Ensure quality data
An analytics application is only as good as the data it uses.
Achieving quality data is a result of several different practices. First, incoming data must be cleaned by purging incomplete or “broken” data; assuring that data records are not duplicates of each other; and making sure that all data is standardized into a single, uniform format, although the original contributing systems may have named and formatted this data differently.
In some cases, this data cleaning and standardization work must be done by hand, but in most instances, there are tools such as ETL (extract, transform, load) software that can do the work automatically, based upon the business rules that companies provide.
SEE: Snowflake data warehouse platform: A cheat sheet (free PDF) (TechRepublic)
Regardless of the tools used (or not used), top-level executives still are not fully comfortable about the quality of their data. In a KPMG study that is now five years old, 56% of CEOs had concerns regarding the integrity of their data. In July of 2021, Gartner said that, Every year, poor data quality costs organizations an average $12.9 million—so not much has changed.
The message for IT is clear: Data quality is still a work in progress, and tools and practices should be in place to assure that the data being used in analytics is of highest quality.
3. Have a data maintenance and tuneup strategy
Twenty years ago, a baseball pitcher’s performance was largely evaluated by his earned run average—the number of runs that were directly attributable to his pitching during a nine-inning game. Now, baseball pitchers are measured by ERA, but also by a slew of new analytics statistics such as K/BB (strikeouts per walks), HR/9 (home runs allowed per nine innings), WHIP (walks over innings pitched) and OOPS (opponent on base plus slugging). The sophistication of sports analytics has prompted the emergence of a new analytics field known as sabermetics.
The data and the analytics that enterprises use are no different. As business and world conditions evolve, how we measure analytics effectiveness must, too. For IT, this means two things:
(bul)The data that analytics operates on should be regularly refreshed to ensure optimal accuracy; and
The algorithms and queries that are used to operate on the data should also be regularly revisited. In other words, are we interrogating data in the most effective ways or should the queries and algorithms that we’re using be revised?
4. Use outcome tracking
IT and end users have a tendency to measure business success in terms of projects completed. Once a project is complete, you move on to the next project in your project backlog.
Moving forward with projects is good—but not if the tradeoff is failing to track the outcomes of those projects completed.
In both the Gartner and KPMG research cited earlier, there were strong indications that CEOs distrusted their data and analytics because they weren’t seeing the direct impact of their analytics on the business. Once they do, trust in the data and the analytics increases.
This is why one of the most important things that IT and end users can do is to track the business track records of their analytics. If the analytics aren’t contributing value to the business, change them or drop them.