I’m going to tell you a story about a Bitcoin NFT marketplace. No company name — NDA — but the lesson matters more than the details anyway.
A bit of context on Ordinals
In early 2023, a developer named Casey Rodarmor did something interesting. He figured out how to inscribe arbitrary data directly into Bitcoin transactions — at the level of individual satoshis, the smallest unit of bitcoin. Images, code, text, music. All of it now permanently stored on the most secure blockchain in the world. No sidechain. No separate token. Just Bitcoin.
That’s how Ordinals were born. And with them, an entire Bitcoin NFT ecosystem.
Sounds simple. In practice — technically complex, fees higher than Ethereum, user experience basically zero. But the market decided it was interesting, and the money followed.
The company I joined
I came into this project as a senior PM with crypto startup experience. The company was building a marketplace for Bitcoin NFTs and had a real technological edge.
The founder — a high-level engineer — had built a system of hierarchical inscriptions with cryptographic proof of provenance. In plain terms: in the Ordinals world, anyone could copy an image, inscribe it on-chain, and claim it was the original. Our technology solved this at the protocol level. Every child inscription carried verifiable proof that it came from a specific parent, and that entire tree was visible on-chain all the way back to the beginning. Nothing like it existed in the ecosystem at the time.
Real technology. Strong team. Marketing had pulled in 50,000 Twitter followers. Everything seemed to be in place.
One problem. We weren’t in production. And the founder didn’t want to go there.
The technology that was supposed to change everything
Let me explain what we actually built — because it matters for understanding why the failure was so painful.
The core problem we solved was provenance. In the early Ordinals world, anyone could take an image, inscribe it on Bitcoin, and claim it was an original. The blockchain stored the data permanently — but it couldn’t tell you whether the inscription was authentic or a copy. This is the fundamental authenticity problem that had plagued Ethereum NFTs for years, now migrating to Bitcoin with none of the infrastructure Ethereum had developed to handle it.
Our solution was a system of hierarchical inscriptions with cryptographic proof of lineage. Here’s how it worked in practice:
A creator would inscribe a single parent inscription — their original work. This parent became the root of an entire collection tree. Every child inscription generated from that parent carried cryptographic proof of the relationship, verifiable on-chain without any oracle or external database. You could trace any inscription all the way back to its origin. The entire lineage was transparent and tamper-proof, directly on Bitcoin.
Nothing like this existed in the Ordinals ecosystem at the time.
But the real innovation was who paid for the inscriptions.
Inscribing on Bitcoin is expensive. Depending on network congestion, a single inscription costs anywhere from 0 to 00 or more. During peak mempool periods — when everyone is competing for block space — costs spike unpredictably. This created a brutal economics problem for creators: to launch a collection of 1,000 pieces, you might need to front 0,000–00,000 in inscription fees before selling a single item.
We inverted this model.
A creator would inscribe only the parent — one transaction, one fee, their cost. The children existed in our database as pending inscriptions: named, described, with artwork uploaded, but not yet on-chain. When a collector arrived and chose a specific piece they wanted, they paid the inscription fee at the moment of purchase. The act of buying was the act of inscribing. The parent-child relationship was established on-chain at that exact moment, with cryptographic proof linking the new child to its parent.
This was lazy minting on Bitcoin — a concept well-understood on Ethereum, now implemented on a blockchain that makes it significantly harder to do.
For creators: the cost barrier to launching a collection dropped from potentially hundreds of thousands of dollars to the cost of a single inscription.
For collectors: they weren’t buying a JPEG in a database. They were triggering the creation of a permanent, provably authentic Bitcoin artifact. Their transaction was the artwork coming into existence.
The economics were elegant. The technology was real. And almost nobody who showed up at launch understood any of it.
The market we were actually entering
Before I get to the internal story, it’s worth being honest about the structural reality of the market we were building in. Because the onboarding failure didn’t happen in a vacuum — it happened inside a market that was tiny, technically hostile, and winner-take-all from day one.
The entire Ordinals ecosystem generated around $725 million in total trading volume by end of 2023 — its first year. That sounds like a lot until you compare it to Ethereum NFTs, which had done multiples of that in a single month at their peak. And that $725M was spread across every marketplace, every collection, every transaction on the chain.
The daily trading volume for Bitcoin NFTs in 2024 averaged around $1-2 million on a normal day. For context: a mid-tier Ethereum NFT drop could do that in an hour.
The user base was tiny and deeply technical. We’re talking about people who self-custody Bitcoin, understand UTXO mechanics, run their own nodes, and follow niche protocol development. In 2023-2024, that was a few tens of thousands of people globally who were both willing and able to trade Ordinals. Not millions. Tens of thousands.
And then there was the marketplace dynamic. Magic Eden — a well-funded, multi-chain platform that had been operating since 2021 — integrated Ordinals in March 2023 and within a year captured over 60% of all Ordinals trading volume. Unisat had first-mover advantage as the technical infrastructure layer. OKX threw exchange-scale resources at the problem. The top three players eventually commanded nearly 100% of volume.
For a small startup with unique technology but no distribution, no existing user base, and no marketing budget at that scale — this was the environment. The addressable market wasn’t “Bitcoin NFT enthusiasts globally.” It was “the subset of that tiny technical audience not already loyal to Magic Eden or Unisat.”
The founder knew this. It was a conscious, accepted risk: we have better technology, we move fast, we capture the early community before the big players wake up. That logic was sound.
The execution — specifically the time-to-market — destroyed the logic before it could play out.
“Users will burn us”
When I joined, the internal conflict was immediately obvious. The co-founder was saying “we urgently need to ship, we’re running out of money.” The founder was saying “we’re not ready yet, the technology has to be right.”
The founder had a history. He’d launched a product before, started listening to users, started building what they asked for — and ended up with something nobody actually wanted. Classic trap: users say one thing, mean another, you build a third. His conclusion was direct: don’t listen to users. Trust the technology, do it right.
I get it. That logic works — in certain contexts.
Here, it was killing the business.
The founder trusted the cloud so little that he physically brought in and installed his own servers for production. Literal hardware. A man of principles — respect. But it was a symptom of a deeper pattern: total control instead of speed, perfectionism instead of feedback.
I spent months pushing the team toward shipping. Finally, we launched.
The launch
Marketing had done its job — 50,000 followers, community buzz, anticipation. For the first sale of our own collection, we were ready to sell a thousand. We sold maybe 150 to 200.
Users showed up and understood nothing.
Not because the product was bad. Because the product was built for people who already knew what an Ordinals-compatible wallet was, how UTXOs work, why fees quoted in satoshis per virtual byte are normal. Those people exist — in the thousands. But the Twitter audience we’d built was regular crypto enthusiasts who’d heard about Bitcoin NFTs and wanted to try.
They landed, saw the interface, had no idea where to start, and left.
Sessions under 90 seconds. Funnel collapsed. The technical users who figured it out loved it. But there weren’t enough of them to build a business on.
What the real problem was
Not the technology. Not the marketing. Not the idea.
The problem was that we spent too long in development without getting feedback from real people. We lived inside our own model of the user — technically fluent, patient, already convinced of the value of Bitcoin NFTs. We drew that person in our heads and built for them.
Nobody sat next to a real human and watched them try to mint. Nobody asked: did you understand you need a separate wallet? Did you understand why you’re paying a fee? Did you understand what any of this actually is?
Not because the team was dumb. Because the founder had made a decision: users shouldn’t influence the product before launch. Wait for the right moment, ship the right thing.
But a “right product” that nobody can use isn’t the right product.
What should have been different
This isn’t an argument for blindly listening to users and building whatever they ask for. The founder was right that this is dangerous.
But there’s a difference between two questions:
“What do you want us to build?” — dangerous. This is where you go off track.
“Can you complete a mint without losing money?” — safe. The answer doesn’t change the product vision at all. But if the answer is no — you have a problem that needs to be fixed before launch, not after.
Three things needed to happen:
A closed beta with explicit success criteria. Not “let’s give it to friends to poke around.” A real beta: 50 people from the target audience, a specific task, a measurable completion rate. If fewer than 60% get through without support — the product isn’t ready, regardless of how elegant the technology is.
Onboarding treated as its own product problem. We treated onboarding as a UI detail. That was wrong. For a product on a new protocol with unfamiliar wallets and new mechanics — onboarding is the product for a new user. It deserved its own design sprint, its own metrics, its own readiness criteria.
Time-to-market as a hard priority. We launched too late. In the Ordinals world, early 2024 was when the community was forming — the first collectors, the first builders, the first social fabric. We were there, but inaccessible. By the time we fixed onboarding, part of that audience had already found others and built loyalty elsewhere.
What I took from this
If you’re not in the market — you can’t do anything. You can theorize, build models, believe in the technology. But you only find out what users actually experience when they try to use the thing.
The founder was afraid users would burn him. In the end, the absence of users burned him instead.
One more thing worth saying: in the Ordinals ecosystem, there’s one player that always wins regardless of which marketplace succeeds or fails. Miners. Every inscription, every transaction, every failed mint attempt — miners collect the fees. By 2023, Bitcoin miners were generating $63 million per day in fee revenue, a significant portion driven by Ordinals activity. The infrastructure layer extracts value independent of product quality, user experience, or who builds the best marketplace.
For everyone else building in this space — the marketplaces, the tooling, the products — it’s a much harder game. Winner-take-all dynamics, tiny addressable market, technically hostile users, and a clock ticking from day one.
The best technology on the market means nothing if people can’t get to it — and especially not if you arrive late to a market that doesn’t wait.
Ship faster. Break earlier. Fix it while you still can.
Be the first to comment