To Improve Data Quality, Stop Playing the Data Telephone Game
Blog: The Tibco Blog
Reading Time: 3 minutes
Do you remember when you used to play the telephone game with other children? You know, the game where the first person in a chain whispers a phrase to the second, then the second repeats it to the third, and on until the last person repeats it back to the first.
It was such a joy to laugh about how what began as “The sun is in the sky” somehow transformed into “It is fun to eat pie” as the phrase passed from one friend to the next.
Further laughs quickly followed when each person in the chain shared their whispered phrase with all, allowing everyone to see what went wrong and where.
The Data Telephone Game
Interestingly, over the past thirty years, data management has adopted the same telephone game formula, copying data from one database to the next, with many stops on the journey.
Take the classic enterprise data warehouse process as an example:
- The data starts as transaction records stored in a transaction system’s database
- Next it moves from the source system to a staging database
- From staging it moves to a data warehouse
- With subsets of that data further advancing for storage within satellite data marts
- Many of which soon feed individual Excel files resident on laptops
Or the more recent cloud data lake paradigm:
- Source data from devices are consolidated on edge databases
- This edge data is then copied into a cloud data lake for further analysis
- Additional data from transaction systems might also be added to the lake
- And to inject historical context, warehouse data might also be copied into the lake
Conceptually, these data management best-practices provide the opportunity to improve data quality by applying selected value-add transformations at various steps. But with so many rigid links in the chain, this data version of the telephone game can often inadvertently turn “sky” into “pie.” The business impact of this quality problem produces anything but childhood chuckles.
How Big is this Repeated Copying of Data Problem?
Just how much data is getting copied? In its Worldwide Global DataSphere Forecast 2019-2023, IDC estimates that for every terabyte of net-new data, over six additional terabytes of copied data is generated via replication and distribution. That is a lot of opportunities for “sky” to become “pie.”
Three Ways to Stop Playing the Telephone Game
IDC’s numbers, when combined with everyone’s telephone game experience, suggest trying a different approach to improve data quality. Here are three common-sense things organizations might consider.
- Copy Less, Virtualize More – Data Virtualization is a proven method for integrating data without physically copying it. This will substantially reduce the transformation errors and entropy inherent in typical multiple-copy, data warehouse, and data lake deployments. Beyond fewer copies, data virtualization directly improves data quality via metadata-driven syntactic and semantic transformations and enrichments that standardize datasets and encourage reuse. Everyone is on the same page. And when things change, as they inevitably do, it’s a lot easier to modify centrally managed metadata definitions than it is to modify multiple distributed ETLs and database schemas.
- Share Reference Data Everywhere – Reference Data Management improves data quality by enabling organizations to consistently manage standard classifications and hierarchies across systems and business lines. This lets them achieve needed consistency and compliance without extra copies. And by adding data virtualization as the distribution method, organizations can easily share and reuse reference data held in one virtual location.
- Think Data Domain, Not Database Technology – Today, there are lots of cool, fit-for-purpose database technologies. But “new and exciting” doesn’t necessarily translate into “high business value.” Instead, think about the most valuable data domains. For example, if customer excellence is your competitive advantage, then focus on improving quality within the customer data domain. Master Data Management is the key to success in this case, allowing organizations to ensure data integrity within selected data domains such as customer, employee, product, and more.
Data Virtualization is a proven method for integrating data without physically copying it. This will substantially reduce the transformation errors and entropy inherent in typical multiple-copy, data warehouse, and data lake deployments.
Click To Tweet
Stop Playing the Data Telephone Game
Let’s leave the telephone game to the kids. Instead, improve your data quality by executing the three common-sense recommendations above with TIBCO Unify. To learn more, talk to TIBCO and our partners.