November 24, 2025 at 12:00 am
Comments posted to this topic are about the item Is Data Modeling Common?
November 24, 2025 at 7:27 am
Hey Steve,
I've been trying a lot of these tools again lately, and DBeaver produces the diagrams that I like most. (And the community edition is free)
It was easy to use, and they just look good.
Regards,
Greg
November 24, 2025 at 8:51 am
Personally, when I design something new, then yes, I produce a logical model. When I inherit something, I'll look at the physical model and do the best I can.
What I would be interested in is a tool that, when given a set of JSON documents for the same type of object, produces an ERD of the inferred schema.
November 24, 2025 at 10:01 am
As a developer, I used Visio for years back when it still included database modeling. Why? Because it was part of the Microsoft suite we were already using—Word, Excel, and so on.
When that feature disappeared, I switched to ModelRight 4.1 for SQL Server as an affordable alternative. I still use it today, probably as the only one on our team. I prefer having a graphical overview because that’s how my brain works.
Newer colleagues rely on the Microsoft stack with Entity Framework migrations. As a result, I often see tables with fields like nvarchar(max) for something as simple as a name. These days, databases don’t seem to matter—until performance issues arise. Then they come looking for the old dinosaur.
November 24, 2025 at 3:52 pm
As a developer, I used Visio for years back when it still included database modeling. Why? Because it was part of the Microsoft suite we were already using—Word, Excel, and so on.
When that feature disappeared, I switched to ModelRight 4.1 for SQL Server as an affordable alternative. I still use it today, probably as the only one on our team. I prefer having a graphical overview because that’s how my brain works.
Newer colleagues rely on the Microsoft stack with Entity Framework migrations. As a result, I often see tables with fields like nvarchar(max) for something as simple as a name. These days, databases don’t seem to matter—until performance issues arise. Then they come looking for the old dinosaur.
Nice, glad to see someone still finding value in modeling.
November 24, 2025 at 3:55 pm
Personally, when I design something new, then yes, I produce a logical model. When I inherit something, I'll look at the physical model and do the best I can.
What I would be interested in is a tool that, when given a set of JSON documents for the same type of object, produces an ERD of the inferred schema.
I'm less concerned about formatting, and I'd like to be able to use a SQL script to generate both a Logical and physical model of what I'm working with. They tend to be pretty close for me, since I mostly see apps on one platform,but it is nice to be able to remove some stuff (indexes, specific types, etc) and focus on how the entities link together.
November 24, 2025 at 3:56 pm
I'm an old hand at data modeling. I started doing them manually in the 80's in the early drawing programs (yeah, majorly tedious!). I was one of the early users of ER for Windows (later ERwin now from Quest) and spent a lot of time with ER/Studio (now from Idera).
One thing many tools leave out is the distinction between a logical model and a physical one. The only ones that handle this properly (that I know of) are Erwin & ER/Studio. Yet, for their price, they should include all the bells & whistles! I'm testing out the new Redgate Data Modeler and I'm hopeful that we'll be able to shape it into a viable, cost-effective alternative to the big boys.
But, to address the question if data modeling is dead, I must emphatically state: No! It has fallen out of use because of the cost of proper tooling. But I think that this new Data Modeler can help bring it back. I find that data teams who maintain the model first have fewer overall inconsistencies in their processes, fewer data anomolies, and are generally able to respond to changes more quickly with less fallout.
LinkedIn: https://www.linkedin.com/in/sqlrv
Website: https://www.sqlrv.com
November 24, 2025 at 8:58 pm
Thanks, I'm hoping you're right, Aaron and cheaper tooling will get more people modeling. I like the Redgate tool, though it needs a few things for me. However, that and other tools are out there at a reasonable cost if people look. They have less features, but most let you update and image and print it out for your team.
November 25, 2025 at 7:55 am
We are using Visio for data modelling, because we have it in the licencing but for the data models we need to create for Power BI from the AWS Glue/Athena databases and for documenting existing SQL Server database content for migrating to AWS.
I first used Visio just before Microsoft bought it and it was a great product with both forward and reverse engineering of databases but Microsoft quickly dropped the forward engineering and made it far less useful and a mere shadow of its former self.
I'd be interested to know about other modelling tools available especially any for working with AWS S3/Glue/Athena datalakes.
November 25, 2025 at 12:09 pm
I'm less concerned about formatting
I'm not concerned about the formatting either. I'm interested in understanding what the JSON documents represent. If I have to ingest JSON documents and their data into a data warehouse, I have to understand the JSON model well enough to model and map its contents to the DW model.
What I find with JSON documents is that the good ones appear to be based on a disciplined object model. The bad ones look like stream of consciousness artefacts (your definition of consciousness may vary).
November 25, 2025 at 10:52 pm
What I find with JSON documents is that the good ones appear to be based on a disciplined object model. The bad ones look like stream of consciousness artefacts (your definition of consciousness may vary).
like XML, TOML, and most any other semi-structured data.
November 26, 2025 at 11:22 am
I'm doing a lot of data modelling, and we're writing and implementing policies that will force around 100 underlying organisations to also start modelling everything they share with other organisations. But tooling is a serious problem.
Unfortunately, Vertabelo and its cousins like dbdiagram.io all have the same issue: nice pictures, and good correspondence to the technical implementation, but pretty much useless for business analysis and modelling purposes. And that is always the hardest part. Technical data models are mainly pictures and programmers are correct to say "but I already have a data model in my code". Business users cannot validate technical models either, even if there is a picture.
The value is in the conceptual model, preferably in the form of verbalizations such as "An order always has at least one orderline." where SMEs can say whether that's correct or not. The only tools I know that do this, are Dutch (CaseTalk and Cogniam) and they have few (but fiercely loyal) customers. We just ran about 60 new trainees through a course where they had to use CaseTalk to build fact-based models and it turns out it is *much* easier to use fact-based (sentence based) tooling for new people than it is for people that already "know how to model" (or at least they think they do).
CaseTalk also transforms those models into correct UML/ERD diagrams, or ontologies, at will, and can forward and reverse engineer them into physical models. However, there is a problem there: you want a DBA to add implementation details to models. So where does that happen? IMO there should be a disconnect between the business and implementation models in terms of the physical concerns being added. You don't want to add tablespaces to logical models.
So Vertabelo and its cousins might just be the wrong tool for most people, because they're built on the technical side by the technical people to solve a few minor technical issues. In the business view, they don't matter - and the business is correct on that, I think. And that is why it seems data modelling is dead: the way it is viewed by the IT community simply doesn't reflect the problems that the business would like to see solved.
November 26, 2025 at 9:24 pm
@Ronald, you have a point that most modeling tools focus only on the physical model. There are some that also claim to have logical models, but they are not true logical models, merely a reflection of the physical. The only tools I've seen that properly handle the distinctions between logical and physical are Erwin or ER/Studio. Even then, they don't tie to conceptual models easily.
Because of these difficulties, it's not surprising to see a decline in data modeling. I truly hope that less expensive tooling with greater functionality will shift the tide back toward data modeling and all the value that it brings. That's why I'm working with Redgate to see if the Vertabelo tool can be enhanced to be more useful and therefore more valuable.
LinkedIn: https://www.linkedin.com/in/sqlrv
Website: https://www.sqlrv.com
Viewing 14 posts - 1 through 14 (of 14 total)
You must be logged in to reply to this topic. Login to reply