Can't Inmon and Kimball just get along?

A longtime data management professional illustrates the synergies that exist between the Inmon and Kimball data warehousing methodologies.

Do a Google search for the phrase "Inmon and Kimball" and it's easy to come away with the perception that one must choose sides between Bill Inmon and Ralph Kimball, the pioneering creators of two very well-known -- if philosophically different -- approaches to data warehousing.

But at least one veteran data warehouse architect -- and Bill Inmon himself -- point out that under the right circumstances, the two ever-evolving methodologies can actually work together nicely.

"They each have their own application," said Bill Harrison, a business intelligence (BI) and data warehouse architect with Omaha Public Power District, a public utility in Nebraska. "Inmon's normalized data model is really good for your centralized data warehouse, but when you go out into your data marts and design those, you want to use the Kimball methodology. So we use both."

A brief history of Inmon and Kimball

Bill Inmon and Ralph Kimball had one key thing in common when they first introduced their methodologies in the early 1990s. They both wanted to help organizations efficiently manage mountains of information and make better business decisions. But they went about it in different ways.

Inmon, known as the "Father of the Data Warehouse," focused his approach on the building of the enterprise data warehouse (EDW) -- a centralized relational database management system that provides users with access to fully cleansed, integrated and normalized data. Inmon is the first to admit that the process of building an EDW is not quick and not cheap, but it can produce a very attractive long-term return on investment.

Kimball, who has been called the "Father of Business Intelligence," championed the creation of data marts, relatively small repositories of information that are designed to support the needs of specific departments within an organization, such as the finance, accounting or human resources groups. While Kimball's dimensional model promises a quick return on investment, experts say maintaining data quality in a sea of data marts can easily become problematic.

The two approaches have evolved over the years, according to Inmon, the author of “DW 2.0: The Architecture for the Next Generation of Data Warehousing.” The Inmon methodology now includes a framework for textual data warehousing and Kimball, author of “The Kimball Group Reader: Relentlessly Practical Tools for Data Warehousing and Business Intelligence”, has highlighted methods for ensuring the consistency of information housed in various data marts.

Kimball could not be reached for comment. Inmon, however, believes that synergies exist between the two approaches.

"The Kimball architecture works really well for data marts for when you have to build out what the end user sees at the marketing, sales, finance or accounting level. That's where the Kimball architecture works," he said. "However, you need to feed the data marts from a centralized, integrated data warehouse -- the classical Inmon-style data warehouse."

Inmon and Kimball work together at Omaha Power

Omaha Power is just one example of the many organizations today that employs both Inmon- style data warehouses and Kimball-style data marts, and Bill Harrison thinks the much-hyped competition between the two methodologies is "a bunch of hooey."

"Supposedly, they've been fighting this big battle for years. It's just a good way to get publicity," said Harrison, who also serves as the president of the Nebraska chapter of DAMA International, an association of data management professionals. "I mean it's ridiculous. They complement each other. You don't use one or the other."

Harrison said the Kimball route is definitely the way to go when building data marts, specifically because the methodology is relatively easy to understand while offering solid performance and quick results.

"The architecture is designed to get the data out [to business users] as fast as you possibly can, and most [source systems] aren't designed that way," he said. "Most systems are designed to get data in as fast as possible."

For example, Omaha Power runs Oracle PeopleSoft to help keep track of its many customers throughout Nebraska. The information tables within the PeopleSoft application are designed to make it easy for users to enter new information.

"But you want a different design for getting data out," he explained. "So, you use data marts and Kimball's methodology, which totally redesigns the data and denormalizes some of it to get it out faster for reporting and queries."

Other times, however, Harrison is happy to have the Inmon methodology working for him, especially when it comes to designing and maintaining the company's centralized EDW. Harrison said it's important to follow Inmon's theory, which promotes the normalization of data, when creating an EDW. This is true regardless of whether the EDW will be used to feed Kimball-style data marts, he added.

"This war has been going on as long as I can remember, but each approach has its own merits," Harrison explained. "So I say: 'Let's use both and get them to work together.'"

Dig Deeper on Data warehouse project management