Sifflet

I coined "data entropy" in 2022 after I'd experienced it firsthand in an environment where bad data = bad P&L = BAD day. Here's the original definition: data entropy is the disorder that accumulates between data and its consumers. In 2022 those consumers were humans. Analysts, dashboards, the CFO squinting at a number. In 2026, the consumers are machine and they have agency. For years we watched data infrastructure evolve in a way that promised more flexibility, interoperability, accessibility and eventually more... entropy. That's why I went back and rewrote the piece from scratch. Updated definition for 2026: Data entropy: the disorder between data and its consumers; humans and agents. Full blog post linked in comments. #dataentropy #datatrust #datagovernance

Four years ago, I gave a name to a problem I'd experienced firsthand. Turns out, I underestimated it.. BIG time. In March 2022, I launched a newsletter called Data Entropy. A year later, I expanded the idea in a longer essay. The argument was simple: data platforms obey their own version of the second law of thermodynamics. Disorder accumulates. Every pipeline, tool, and transformation creates another opportunity for meaning to drift. Left alone, entropy only grows. At the time, I thought it was a tax. Annoying, measurable, manageable. Something better tooling would eventually contain. And it did, to an extent. Four years later, the problem only got worse despite "better tooling". Teams have more monitoring, testing, and cataloging than ever before. Yet when I ask data leaders whether they trust the number on the dashboard right now, the honest answer is usually no. Entropy wasn't contained. It compounded. And something has changed that makes me realize the 2022 definition no longer explains what we're seeing. I've started writing about it again. More on Monday. The AI generated picture is intentional ;-) #dataentropy #datatrust #ai #datagovernance

Thousands of enterprise data teams analyzed. We have looked at what is actually shipping, breaking, and getting fixed. Most of it points in the same direction. A 2026 outlook on where data infrastructure is heading: • Metadata becomes the control plane • Observability won’t be bundled • AI fails without reliable data 7 shifts shaping how data teams operate next → Download the report

Data leaders know this already: Things will break. Not because your stack is complex - but because it’s fragmented, constantly changing, and quietly failing underneath. A schema change. A missing value. A dependency nobody tracked. Most teams don’t catch it until it reaches the business. That’s what’s changing in 2026. 7 predictions on how data leaders are detecting issues earlier - and understanding impact before it spreads →

Data pipelines break, but that's not the problem. The problem is your engineers spending entire days figuring out why. While stakeholders lose faith in the numbers. While decisions get made on gut feeling. Callum O'Connor, Senior Analytics Engineer at Adaptavist, lived this. And he was kind enough to talk about it on camera. What changed with Sifflet: resolution time went from hours (sometimes days) to minutes. Stakeholders started trusting the data again. The team got their days back. Not because of better alerts. Because they finally had lineage, business context, and real visibility into what was happening and why. What we call the Control Plane for Data & AI. But the part we liked the most was this: "It didn't feel like we were customers. It felt like we were partners." That's the thing we care about most at Sifflet. Not just the product working. The people behind it feeling supported. Watch Callum's story here 👇

Data pipelines break, but that's not the problem. The problem is your engineers spending entire days figuring out why. While stakeholders lose faith in the numbers. While decisions get made on gut feeling. Callum O'Connor, Senior Analytics Engineer at Adaptavist, lived this. And he was kind enough to talk about it on camera. What changed with Sifflet: resolution time went from hours (sometimes days) to minutes. Stakeholders started trusting the data again. The team got their days back. Not because of better alerts. Because they finally had lineage, business context, and real visibility into what was happening and why. What we call the Control Plane for Data & AI. But the part we liked the most was this: "It didn't feel like we were customers. It felt like we were partners." That's the thing we care about most at Sifflet. Not just the product working. The people behind it feeling supported. Watch Callum's story here 👇