Lessons from building data systems in environments where a wrong number has a real cost.
I've never vibe-coded or agentically engineered. Only directed engineering. Here's why, and what it means for real data infrastructure.
A packaging manufacturer lost power for a fraction of a second at a time. It took six weeks to find out why.
I recently migrated a large, closely watched metrics environment off a BI semantic model onto a cloud warehouse. I architected both ends. They still didn't reconcile perfectly, and I had to document every difference and make it acceptable to leadership. That experience is why I take metric migration seriously.
ERP exports break in predictable ways. Here's how the OIM handles the full range of it. From encoding errors at the gate to null values that make it all the way to the dimension table.
A six-layer architecture that never needs its SQL touched. Here's how config-driven design, dimensional modeling, and a parametric targeting engine converge.
I picked up The Goal trying to get better at influencing executives. What I found instead was a different way of thinking about operations entirely.
25 runs. 76 million rows. Zero failures. How I built the adversarial test that earns the right to call the OIM production-ready.
I knew Power BI better than almost anyone. That's exactly why I walked away from it.
2.5M rows was the proof of concept. 50M rows in under two hours on a laptop was the proof it was real.
The OIM started as a Power BI template. Power Query was the plan. Then I hit a wall and found the tool that changed everything.
Most manufacturers are sitting on years of operational data they never use. Here's what it actually takes to turn that into a clear picture of your floor.
The reporting pattern I earned trust in during the first migration had to hold up again, this time in real time, tracking thousands of assets against a moving target. Here's what I learned.
A hurricane exposed a leadership reporting tool held together by macros. I said I'd figure it out. Eight weeks later I had a working replatform. It took months more to earn the trust that made it real.
IT said 4-6 weeks. My executive review was in two. So I figured it out myself. That's where the obsession started.
Build for the person doing the work. Everything else follows.
Nobody cares how complicated your analysis was. Here's the one thing that actually gets acted on.
One operator, 35 work orders, and a shift report that still couldn't answer a simple question. The story behind the OIM.
Dashboard graveyards exist for one reason: they don't answer the questions that drive action. Here's the test for whether yours will get used.
Most analytics tell you what happened. The ones that move organizations tell you what to do next and make sure everyone is looking at the same thing.
Why knowing what to work on next is the only thing that actually moves the needle.
345 Intelligence facilitates the full lifecycle of data trust for manufacturing operations. The name carries two metaphors, both structural, both rooted in infrastructure that operates beneath notice but bears the weight of everything built on top of it.
Practical takes on manufacturing data, ERP analytics, and the real work of turning messy source systems into something you can trust.
You're in — first issue lands soon.
Please enter a valid email address.