The Insanity of Data Education
The Weekend Windup #22 - Reflections, Cool Reads, Events, and More
Yesterday, I published a piece on the organizational crisis in data modeling, based on a survey of over 1,100 data professionals. The overarching theme? A whopping 89% of respondents are struggling with their data modeling approach. Only 11% say things are actually going well.
When you dig into the complaints, the numbers paint a bleak picture:
59% cite the relentless pressure to move fast.
51% suffer from a complete lack of clear ownership.
The TL;DR here? It’s your boss’s fault. I’m half-joking, but not really. More like your boss’s boss’s boss’s fault.
Organizations simply do not give data practices the time, attention, or agency they deserve. Data teams are perpetually caught in a game of “not it” when it comes to ownership, resulting in a relentless cycle of firefighting. As Jack Bergman wrote, “There’s never enough time to do it right, but there’s always enough time to do it over.”
But the survey data isn’t actually what I want to talk about today. I want to talk about some recent reactions to it.
The 40-Year Echo Chamber
When I shared these findings on LinkedIn yesterday, the comments section lit up. A friend of mine chimed in, lamenting that practitioners are completely confusing conceptual and logical data modeling with physical schema design. His take essentially boiled down to: It’s annoying that we’ve been doing data management for 40 years and people still don’t have a good handle on what data modeling is or why we should be doing it.
He’s technically right about the definitions. But I had to call it out for what it is: condescending.
If the data industry has been teaching and preaching the same way for four decades and the vast majority of practitioners are still struggling, clearly the approach hasn’t worked. At this point, we cannot keep blaming the practitioners. Beating people over the head and telling them they’re stupid for not adhering to 40-year-old practices, or dogmatically blaming them for not rigidly following data modeling methodologies is unhelpful at best, and gatekeeping at worst.
Given a lack of time to do things “right”, people will just ignore the “best practices” and find workarounds instead. From what I can tell, this isn’t unique to our world. Yesterday, I was listening to an AI Daily Brief podcast, where the same problem was arising with learning to use AI tools in the workplace. People want to learn the best ways to use these tools, but because they are under extreme time pressure to deliver with AI, they ironically don’t have the time to learn to use them properly. Any investment in learning must take place outside of work, on one’s own dime.
Doing the same thing over and over again and expecting different results is the literal definition of insanity. Yet, the data industry keeps insisting that if we just shout louder at people, they’ll finally get it.
Stop Blaming the Victims of Dysfunction
I travel the world talking to data practitioners and leaders. Overwhelmingly, these are highly intelligent people with great intentions. They’re not dumb or lazy. Far from it. But they are operating with a severe lack of agency. Often, they feel they don’t have control over their time or their careers.
When you are pressured for time and just need to get something out the door - because it’s always about getting something out the door - reading dry, preachy textbooks on relational theory isn’t going to save you. Some of the classic materials out there are drier than the Sahara, unapproachable, and frankly, completely detached from the organizational dysfunction most engineers face daily.
Part of the problem is a disconnect between the ideals we hold and the realities of where practitioners operate. If you want people to implement good data governance or modeling under immense time constraints, you have to meet them where they are.
I was recently talking with a fellow educator who told me they “don’t want to spoon-feed information.” On one hand, I get it: make the students work hard. But it’s also an incredibly lazy and arrogant take. It seems like it’s almost a badge of honor for some educators to overcomplicate things, use complex jargon, tell people to “just read the docs,” and ridicule them when they can’t keep up. I’ve had professors like this, and I can’t say their approach was effective. But it taught me the type of educator I want to avoid being.
Another person replied to my LinkedIn post today, blaming organizations’ failures on people’s lack of knowledge and skills. It’s easy to blame people for failing to learn skills and knowledge. But the survey data shows that many people are under immense time pressures. Learning under duress is incredibly difficult, especially if you have a family and responsibilities outside of work.
If you are trying to teach someone, your literal job is to figure out the messaging that makes it click for them. You have to give them the building blocks so they can navigate their specific, messy reality. If we want to stop repeating the same mistakes, we have to change our approach.
Having spent years writing books and articles, podcasting, building curricula, advising universities and educators, and teaching data concepts to massive global audiences, I’ve learned one universal truth: never underestimate how little people know about a topic. Make the material almost absurdly digestible, approachable, and maybe even a little fun.
How We Actually Break the Cycle
If we want to stop having the exact same conversations 40 years from now, the industry needs a massive reset on how it approaches education and implementation. Here is how we start:
Teach Building Blocks, Not Religion. We have to drop the dogmatic, 600-page academic textbooks. Those have a purpose, but they’re also daunting for beginners. When someone’s hair is on fire, they don’t need a lecture on the chemical composition of fire; they need a bucket of water. We need to teach pragmatic, modular primitives that practitioners can actually use to survive their day-to-day, without treating them like idiots for not building perfect data models, ontologies, etc.
Invest in Your Team’s Growth. Invest in Your Growth Too. If you’re giving people powerful tools (AI, data stacks, etc) to solve real problems for your organization, for God’s sake, invest in their ability to use these tools to their fullest capability. For example, I’m teaching my son to drive right now. Do you think I’d just hand my kid the car keys, not teach him how to drive or traffic laws, and tell him to “figure it out. Go to the grocery store while you’re at it.” Of course not! But that’s exactly how I see quite a few leaders treat their team’s education and training. Of course, this also means people need to be proactive about their own development. Especially these days, when things are moving at warp speed, you need to stay on top of things, learn the fundamentals, and dive into the latest tools. Complacency is a death sentence for one’s career.
Solve the Ownership Void. The survey data didn’t just highlight time pressure; it highlighted a 51% lack of ownership. Education is completely useless if people are stuck in a perpetual game of “not it.” Leaders need to establish clear boundaries, assign actual ownership, and give practitioners the top-down air cover they need to pause, learn, and build things right.
Compete for Attention. Let’s be honest about the modern workplace. If your documentation, training, or book is boring AF, it’s going to lose to ChatGPT, Slack pings, endless meetings, and doomscrolling. If you are asking for someone’s severely limited time, you have an obligation to make the material engaging, digestible, and directly applicable to their pain.
This is exactly why I’m taking a completely different approach with my upcoming writing on practical data modeling and the concept of “Mixed Model Arts.” This isn’t meant as a shameless plug (but please like and subscribe ;).
Lord knows when I started this project, I got my share of haters. “Who is Joe Reis to write about data modeling?! Burn the heretic!” Much of the criticism was from data model’s old guard, as well as a few gremlins living in their mom’s basement. Well, fast forward a couple of years, and as far as I can tell, it’s already one of the most popular resources in the world for learning data modeling.
I took that hate to heart, and realized I didn’t want to be like them. I am not here to scold you from a pedestal or preach 40-year-old gospel. The goal is to meet you exactly where you are, strip away the gatekeeping, and hand you the building blocks. I want to give you the tools to make your own decisions for your specific, messy organizational reality. Hopefully, I can make it a little fun to read along the way.
The industry needs to wake up. Let’s get our act together, fix the messaging, and finally start making data education work for the people actually doing the heavy lifting.
Check out the 2026 Practical Data Community State of Data Engineering: Interactive analysis tool + dataset
The survey has an alternative super corporate BI tool, complete with 3D donut charts and speedometer charts, exports to Lotus and Crystal Reports, and more. Cutting edge BI tooling at its finest.
Also, listen to this as a podcast. Available on Spotify, Apple, and wherever else you get your podcasts. Please support the show with a review. It means a lot.
Have a great weekend,
Joe
Awesome Upcoming Events
All technical. All data. All day.
On March 26, 2026, I’m teaming up with Confluent for a single-track, deeply technical data engineering event in San Francisco.
Capped at 100 practitioners.
No salespeople. No sponsors. Just talks + the hallway track, followed by happy hour.
When: March 26, 2026 · 10am–4pm + happy hour
Where: L4, 945 Market St, San Francisco
Capacity: first 100 data engineers/architects
Thanks to Confluent for partnering on this event and newsletter.
Some friends of mine are doing these events:
SLC
Kyle Nesbit (CEO of Credible, ex-Google) will talk about Giving Data Value to AI at the Utah MLOps Meetup on Tuesday, February 24th.
Mountain View
My good friend Demetrios and the ML Ops community are doing an amazing event on Tuesday, March 3rd in Mountain View, at the Computer History Museum.
Coding Agents: AI Driven Dev Conference.
But wait, there’s more!
Cool Videos and Reads
Why are we still using row-based protocols like ODBC and JDBC in a column-oriented world? In this episode, I sit down with Ian Cook, co-founder of Columnar and a long-time Apache Arrow contributor, to discuss the critical infrastructure changes needed to speed up modern analytics and AI.
We dive deep into the technical bottlenecks of legacy standards - specifically the "serialization tax" of converting columns to rows and back again - and how ADBC (Arrow Database Connectivity) solves this by keeping data columnar from end-to-end. Ian also shares his insights on the intersection of tabular data and LLMs, why AI agents need better access to OLAP systems, and the tension between vibe coding speed and the stability required for critical open-source infrastructure.
Here are some things I read this week that you might enjoy.
The Mythical Agent-Month – Wes McKinney
AI ‘Arms Race’ Risks Human Extinction, Warns Top Computing Expert - Barron’s
Chatbots Are the New Influencers Brands Must Woo - The New York Times
The Worst-Case Future for White-Collar Workers
A chat with Byron Cook on automated reasoning and trust in AI systems | All Things Distributed
How Data Governance Is ActivatedWhat Matters Most in Life?
The Era of the Mixed Model Artist
Find My Other Content Here
📺 YouTube - Interviews, tutorials, product reviews, rants, and more.
🎙️ Podcasts - Listen on Spotify or wherever you get your podcasts
📝 Practical Data Modeling - This is where I’m writing my upcoming book, Mixed Model Arts, mostly in public. Free and paid content.
The Practical Data Community
The Practical Data Community is a place for candid, vendor-free conversations about all things tech, data, and AI. We host regular events such as book clubs, lunch-and-learns, Data Therapy, and more.
One of my goals as a leader for data product is that my team sets stakeholder expectations on timeline to buy the engineers enough time to model, write tests, and not context switch constantly to address crises created by the business not planning and bringing us emergency requests. It's not always easy or fun but it pays off.
I've also been introducing my team to conceptual modeling - we watched some YouTube videos and articles, discussed in a team meeting and one of my team members talked about applying it retroactively to a big project they had just finished to apply the learnings. It was so fun. We're looking at piloting Ellie.ai as a place to practice this and connect it to the engineers' data modeling process next. I think learning in a group + sharing learnings/applications of the work is really motivating.
It's really hard to find time to read work books. I either have to do it on audio or in very small snippets of time (if there are diagrams, etc). Excited to see how you approached writing this book.
Bingo! Computer science education has always been bad. Many instructors adhere to the “You can’t teach programming” paradigm. Which is wrong.
Looking forward to your book!