Data Modeling with Rich Fox
Rich Fox, analytics consultant and educator, discusses the ins and outs of data modeling, including why it’s still an essential practice in the age of AI.
- By Upside Staff
- March 22, 2024
In this “Speaking of Data” podcast, Rich Fox discusses the ins and outs of data modeling and why it’s still an essential practice in the age of AI. Fox has almost 25 years of experience with data and analytics, both as practitioner and as a consultant and educator, including as a TDWI faculty member. [Editor’s note: Speaker quotations have been edited for length and clarity.]
“I come from the business side,” Fox said, “so it’s still a matter of answering business questions. That takes data. As I tell people, I can do all kinds of interesting analytics with data, but if I have no data, there’s nothing I can do -- so the data is critical.” He explained that his definition of data modeling is simply bringing together disparate data sources from both inside and outside the company and combining them for business and analytics purposes.
Fox pointed out that the main recent change has been the increase in the use of external, third-party data to blend with internal first-party data. “The reason this is important is because much of the data on sources systems isn’t clean,” he said. “That doesn’t mean data that’s inaccurate, but rather data that doesn’t align with fields from other sources.
“A great example of that is dates. Some sources may have it month/day/year where others may have it day/month/year and you can’t just merge those. Some sources may have leading zeroes for their customer ID fields where others don’t. We run across those things all the time, so one of the main purposes of data modeling is to create a multidimensional data model for different purposes, including analytics.”
Fox noted that there are always new competitive threats to the business that need to be met, usually with some new data source. He gave the example of his work for a restaurant group in 2020 when COVID-19 hit. The company wound up acquiring epidemiological data from Johns Hopkins to assess the impact of the pandemic on their 750 restaurants and to determine how to best respond, down to a particular locality.
He added that many companies still struggle with modeling their data, especially with so many organizations moving to the cloud. “Many companies, when they moved to the cloud, just lifted and shifted their existing data warehouse architectures,” he said. “They didn’t take the opportunity to optimize their models for the cloud. Now they want to revisit their models to upgrade them and take advantage of the cloud technology.”
When asked about the potential impact of generative AI on data modeling, Fox was reserved. “AI could have the same ramifications as the Industrial Revolution had 100 years ago, but I don’t think it’s going to happen quickly,” he said. “I was fortunate enough to work on a team about 12 years ago that was rolling out IBM’s Watson -- the AI that had won Jeopardy -- and I was able to see the behind-the-scenes operations. One of the areas they were very focused on was in healthcare, to see how Watson could help doctors accelerate the diagnosis of cancer or precancerous conditions. What we found was that, even though the promise was tremendous, it took years to develop and train the algorithms.”
He explained that AI will have a significant impact on how business is done and the day-to-day operations and processes. After all, he pointed out, the cost of doing business keeps going up, but businesses can only raise their prices so far before customers go elsewhere. As a result, he said, the only solution is to make your operations more efficient and less expensive, and that’s where improvements such as AI and robotics can help.
For more information about Rich’s upcoming TDWI data modeling seminar (April 8-10, 2024), visit TWDI.org.