The First Businesses That Accepted It, and Why Most Stopped

The WordPress.com announcement from 15 November 2012 is still live. It is titled “Pay Another Way: Bitcoin”, it explains why the company was adding the option, and it names the processor handling it. What it does not carry, reading it now, is any note, strikethrough or editor’s addition saying whether the option is still there.

That asymmetry is the whole subject. The start of acceptance is a dated, permanent, indexed document. The end of acceptance is a checkout page quietly losing a radio button. One of those is retrievable years later and the other is not, and every count of merchants ever published was built on the retrievable half.

June 2011: an organisation stops, and says why

The earliest well documented reversal is a good place to start because it is unusually explicit about its reasoning.

On 20 June 2011 the Electronic Frontier Foundation published a post removing its Bitcoin donation option. It gave three reasons, and all three are operational rather than political. On the legal position: “We don’t fully understand the complex legal issues involved with creating a new currency system.” On the donors: “We don’t want to mislead our donors”, with the specific discomfort being that the organisation was not comfortable spending what it had accumulated. And on being read as a recommendation: “People were misconstruing our acceptance of Bitcoins as an endorsement of Bitcoin.”

Read the second reason again, because it is the one that generalises. An entity that accepts something it is unwilling to spend has not added a payment method. It has added a holding, and holding was not what it signed up for. That problem belongs to any recipient whose costs are denominated in something else, which is very nearly all of them.

What “accepts” was actually describing in 2011

Before processors existed, accepting meant publishing an address, watching for confirmations, and doing whatever conversion you needed yourself. Everything that follows from that is on the seller. The exposure between quoting a price and settling is on the seller. Reversing a sale, which the protocol has no concept of, is a manual transfer the seller has to initiate and cannot compel.

Sellers of digital goods with no returns and no support obligations could carry that. Sellers of physical goods largely could not, which is why the earliest sustained acceptance clustered where the alternative was unavailable rather than where the option was better.

The price itself was the other half of the problem. As how a thing with no price gets one sets out, arriving at a quotable rate was a construction, and a seller quoting in a unit whose reference rate was itself young is taking a position whether or not they intended to.

November 2012: the same word starts meaning something else

The WordPress.com post names BitPay as the payment processing partner. That single fact changes what the verb is describing.

With a processor in the path, the merchant is quoted a fixed amount in local currency, the processor takes the incoming payment, and the merchant is settled in the currency their costs are in. The merchant does not hold the asset, does not watch confirmations and does not carry the exposure between quote and settlement. The processor does, and prices that service.

The stated reason in that post is worth quoting because it is not the reason usually attributed to merchant adoption. It was about who could pay at all: “PayPal alone blocks access from over 60 countries, and many credit card companies have similar restrictions”, and therefore “we don’t think an individual blogger from Haiti, Ethiopia, or Kenya should have diminished access to the blogosphere because of payment issues they can’t control.” That is a coverage argument. It says nothing about fees, speed or ideology.

May 2013: the reversal reverses, for the mechanical reason

On 17 May 2013 the EFF announced it would accept Bitcoin donations again. The post is direct about the earlier decision: “Two years ago, EFF decided to stop taking Bitcoins for a number of reasons and returned the coins to the community via the Bitcoin faucet.” The coins going back through a faucet connects this directly to the distribution problem, which is a small detail and a real one.

What had changed was infrastructure. The reason given for resuming was the availability of a processor that converts donations automatically, which removes the accumulation problem that reason two had been about in 2011.

Line those three dates up and the shape is clear. The 2011 stop and the 2013 restart are the same organisation with the same concerns, and what moved in between was not the asset. It was that somebody else had agreed to hold the exposure for a fee. Acceptance in 2013 was a different arrangement wearing the 2011 word.

Why the count is not worth having

At this point the metric collapses, and it is worth being precise about why rather than just declining to quote it.

A number of merchants accepting is almost always a processor’s count of its own merchant accounts. That is a vendor reporting its customer list, which is a fine thing for the vendor to publish and is not independent evidence about anything. It counts accounts rather than sales, it counts an enabled checkout option the same as a used one, and the party producing it is the party whose business grows with the number. That does not make it dishonest. It makes it unusable as a measure of adoption.

The alternative source, counting announcements, is worse, because the announcing and the un-announcing are not symmetric. This piece therefore states no figure for how many businesses accepted at any point, and names no business as currently accepting, because current status is a live fact that would have to be checked against a live checkout rather than against a document.

Expanding, contracting, and nobody watching either

This is a genuine crossing in both directions, and it is the clearest case on the site of a metric that moves in one direction loudly and the other direction in silence.

What was expanding was the count of announcements, and along with it the infrastructure that made acceptance survivable for an ordinary business: a processor, a settlement path, a fixed quote. What was contracting, over the same stretch, was the share of that acceptance that involved the merchant touching the asset at all. Acceptance got easier and thinner simultaneously, and both changes came from the same intermediary.

Who could tell at the time is the answer that indicts the metric. The people who could tell were the processors, because they were the only party who saw both the account list and the transaction volume across it, and they had no reason to publish the ratio. Everyone outside had the announcements. A merchant who switched off the option after a quiet quarter wrote nothing, told no one, and left no artefact, and that merchant is invisible in every retrospective count including this one.

That is the same failure mode as the pizza transaction, from the other end. One transaction survives because somebody wrote it down. Thousands of merchant decisions do not survive because nobody did. The record is not a sample of what happened. It is a sample of what somebody had a reason to publish.