This is part 4 in the Field Notes on Perception series, where I’m exploring how we see structure—inside data systems, inside organizations, and inside ourselves. This one’s about the moment when the word “metadata” makes everything feel too complicated—and why that feeling isn’t failure, but invitation. If you’re new here, Part 1 is All Code Is Pressure, and Part 3 is Why I Voted for Command Reveal.
Say the word “metadata” in a meeting or a demo and you can watch people’s eyes glaze over in real time. Trust me, I have seen it 100 times. It’s one of those words that’s somehow too abstract and too jargony. A word that sounds like it came out of a failed ontology startup.1 A word that gets nodded at politely and ignored structurally.
People don’t just not understand it. They lose the plot completely. And maybe it’s because Zuckerberg made “meta” mean everything. And since data is everything, “metadata” starts to feel like everything about everything—which, of course, means nothing at all.
But the problem isn’t just the word. The problem is that we’re trying to explain metadata in a world that still misunderstands data. We act like data is a reflection of reality. But most of the time, it's a stand-in. A proxy. A representation. 2
That row in Excel isn't a loan.
That dashboard isn't the market.
That metric isn't the behavior.
We treat these second-order objects like facts—but they're fossils. Artifacts shaped by systems, incentives, assumptions, and language.
Let’s make that concrete. Take a row in a customer database:
There’s a “status” field marked ACTIVE.
There’s a “last login” from 8 months ago.
There’s a “subscription_tier” set to Premium.
The “billing_status” shows failed payment.
Is this customer active?
To the product team, maybe—they still have a premium flag. To finance, no—the payment failed. To marketing, yes—they’re still on the newsletter list. To support, probably not—they haven’t logged in since last quarter. To the Founder, of course—everyone wants to be our customer!
The row didn’t lie. But it didn’t clarify either. It reflects the assumptions of whoever defined “active.” And everyone brings their own baggage to that party.
Now multiply this tension across dozens of systems and teams—and you see why alignment isn’t just a technical challenge. It’s a human one. And this is one of the pieces of data we care about the *most*. If we struggle with this one, what does that mean for the next hundred tables\columns? (And do we even have the mental capacity or the time to think about and discuss 100 columns?)
On top of all of this, now we introduce metadata. Which is just data about data. Or put another way:
A description of the fossil, not the living thing it once pointed to.
And that’s where the mental collapse happens.
Because if your relationship to data is already fragile, then adding a third layer of abstraction—to something you already don’t fully trust—breaks your ability to reason through it.
So folks check out, they stop processing, and they move on.
But here’s the turn:
Metadata isn't the abstraction. It's the anchor.
Because while data is often conceptual, metadata is experiential.
You might argue about what a row means.
But there's no argument about how many rows are in the file.
Or when it was last modified.
Or who last changed it.
Metadata is:
Who touched it
When they touched it
What system it came from
How long it lived
What else it connected to
It’s also:
What column names changed last quarter
Who changed the logic in the “active user” definition
What upstream source stopped populating last week
What internal system owns the truth for that field (and whether it still exists)
Who uses which tables and how
It's structure, not interpretation. And structure can be traced.
In Why I Voted for Command Reveal, I wrote about how technology at its best isn’t just automation—it’s revelation. About how our systems should preserve judgment, friction, and attention, not erase it in the name of efficiency.
🧪 Why the name Command Reveal
This wasn’t a late-night domain grab. We did the full branding exercise. Sat with it. Dug deep. Brought in outside help.
Judgment is when a data engineer stares at two conflicting “active user” definitions—one from marketing, one from product—and instead of reconciling them in a new field, asks: What are we actually trying to understand here?
Not “how do I model this,” but who are we accountable to, and what happens if we get this wrong?
It's the moment when engineering stops being plumbing and starts being art.
Friction is when a pipeline fails because an upstream field was renamed, and it was designed to fail.
No fallback. No silent defaults. Just a hard stop because someone had the courage to make breakage visible—to signal that something important changed, and it’s worth noticing before trust erodes downstream.
It’s a refusal to let convenience outrun clarity.
Attention is when a dashboard tells you not just the number, but where it came from.
Who touched the logic. What assumptions it rests on. How many hops it took to arrive at this visual.
It’s the rare moment when a chart doesn’t just show you data—it shows you its lineage, its tension, its cost and who you should call to discuss.
And that changes how you use it.
Metadata is part of this posture. It asks something of you. It’s not convenient. It doesn’t flow. But it lets you trace where you are, how you got here, and what shaped the terrain.
It’s not extra.
It’s not optional.
It’s the signal that makes the rest make sense.
And yet—most data conversations break down before we even get to the signal. Because we’re not just missing metadata. We’re missing a shared understanding of what the data ever meant to begin with.
Because what looks like a technical disagreement is usually a semantic one (almost always)3. We think we’re debating facts, but we’re actually debating definitions. We think we're aligning on “active user” when everyone's version is shaped by different systems, experiences, and priorities.4
Here’s what that might look like:
Marketing counts anyone who opened an email in the last 90 days
Product counts anyone who logged in twice this week
Engineering is pulling from a table where the “active_flag” is updated nightly—unless the job failed
The CEO will take whichever one is highest.
The Investors prefer whichever one is lowest.
Data isn’t just misused by accident.Sometimes it’s twisted on purpose—cherry-picked, redefined, quietly shifted to fit the story.
But far more often, it’s misused because no one ever agreed on what it was supposed to mean in the first place.
And in that fog, metadata becomes a flashlight.
So what does that mean in practice?
It means the metadata isn't extra. It isn't exhaust. It isn't "just for compliance."
It means metadata is the thing you always reach for when the data stops making sense (every time).
It tells you:
Where this came from
Who shaped it
What else it touches
How long it's been in play
And most importantly:
Whether you should trust it right now, for this use case, in this context.
Here’s what I’ve learned:
If you treat metadata as an afterthought, your data will never feel grounded. And your team will never stop arguing over whose definition is “right.”
But if you treat metadata as a first-class object, the fog lifts. You can trace decisions. You can audit change. You can spot pressure. You can build trust.
Not by declaring truth. But by revealing the path truth took to get here.
So the question becomes:
Are you chasing the right answer—or tracing how the answer was made?
Here’s what I do:
The word metadata isn’t nearly the only thing that breaks my brain. Almost everything does. In 2025, there’s no single simple truth that isn’t wrapped in a hundred layers of interpretation, transformation, and context. The veil doesn’t lift anymore, it thickens.
So I stopped looking for meaning when something feels too complicated. I stopped talking about it, I stopped thinking about it and I started practicing. Metadata first.
Happy Wednesday!!
Field Notes on Perception is an ongoing series about how we see—systems, data, patterns, and the invisible structures that shape how we think and build. It’s a personal lens on trust, clarity, and cognition in a noisy world. Part 3 is Why I Voted for Command Reveal. Part 2 is Seeing Structure. Part 1 is All Code is Pressure.
The one I always remember is Ontocalypse Productions from Kelly Chase of UFO Rabbit Hole podcast fame. I think its the combo of ontological shock and apocalypse and for some reason its so hard to pronounce on first glance it has always stuck with me.
As in full-on Guy Debord, The Society of the Spectacle representation. Full transparency: I listen to the audiobook once a year. There’s something about the numbered structure, the tone, and the rhythm of the narration that I find weirdly soothing. ChatGPT told me: it’s also… kind of on-brand that you’d find a voice reading numbered theses on mediated reality “soothing.”
A little on data conversations from a guest post about metadata last year: https://more-than-numbers.count.co/i/142292525/the-need-for-a-common-language.
So much more on this in subsequent posts. Relevance sorting, social conformity, language games, the philosophical well is so deep here - hoping I can do it justice.