AMA — four years running no-KYC cards at volume, ask what you like

Do you have decline data broken out by merchant category? That's the number I've never been able to get anyone to publish and it's the one that would actually change how I set up.

I do, with the caveat that it is my merchant mix and not the world's.

All no-KYC cards pooled, 2024 through May 2025, 1,847 authorisation attempts:

Category	Attempts	Declined	Rate
Hosting / infra	612	19	3.1%
Ad platforms	488	71	14.5%
SaaS / tooling	501	44	8.8%
Data / lists	141	27	19.1%
Everything else	105	6	5.7%

The data row is small enough that I would not bet on 19.1% being the true number, but the ordering has been stable for three years and matches what I would have guessed. Hosting companies want your money and ask no questions. Ad platforms are running risk models that specifically look for this.

The number that surprised me when I first built this table was SaaS at 8.8% — I would have said 3% from memory, and that gap between what I remembered and what I had recorded is the reason I keep records at all.

Four years, eleven providers, $640k, and a public post enumerating your spend categories, volumes and dates.

What is your reasoning for posting this? Not a rhetorical question. You clearly think about risk carefully in every other part of your setup, so I assume there is a model behind deciding this thread is safe, and I would like to know what it is.

There is one and it may be wrong.

The model is that aggregate figures across four years and eleven providers, with no provider named against any specific transaction, no dates finer than a month for anything recent, and no chain or address information, do not narrow the field usefully. The population of people doing roughly this at roughly this volume is in the thousands. I have not posted anything that gets you from the thread to me.

What I have posted that I would not post again: the 14 July 2021 date in the opening post. That is a specific day tied to a specific first transaction, and if anyone ever holds a 2021 database from a dead provider, that line is a join key. It was a rhetorical flourish and it was careless.

The reason I'm doing it at all is that the alternative is what the board currently has, which is four hundred anecdotes and no base rates. Someone has to post numbers or the whole category runs on vibes.

That is a better answer than I expected and the July date is exactly the line I would have flagged.

No further objection. The base-rate argument is right.

Leaving it up rather than editing, since an edited opening post with a gap in it is its own signal.

which chain do you fund on and has that changed over four years

Changed twice and both times for cost rather than anything cleverer.

2021 to mid-2022 was mostly BTC, because that was what the providers took and I did not know better. Network fees on top-ups in that period cost me about 1.4% on average on top of whatever the provider charged, which nobody was counting at the time and which was frequently larger than the provider's fee.

Mid-2022 to 2024, ERC-20 USDT. Objectively worse on fees for most of it. I moved because two providers stopped taking BTC and I did not want two funding flows.

2024 onward, Tron. Costs somewhere between forty cents and a dollar a top-up and I have not thought about network fees since, which is worth more than the saving.

If you are asking whether the chain affects acceptance or speed on the card side: not in my data. Confirmation counts vary by provider but every one of them credits on the same schedule regardless of which chain it came in on.

In four year did you ever get money back from a provider after they lock it? Not the ones who die, i mean the ones still alive who just hold your balance and ask question.

Caveat, this may be outside what you track.

Have you measured whether issuing many cards from one account changes anything downstream — decline rates, review frequency, anything? I run one card per buying account and I've always assumed it's invisible to the provider, but I've never had a way to check.

Partially, and the answer is unsatisfying.

I have two accounts with the same provider, one that has issued 60-odd cards over two years and one that has issued four. Decline rates across them are 7.9% and 8.4% respectively, which is a difference of nothing at those sample sizes.

What I have not been able to measure is review frequency, because reviews are rare enough that I have maybe six events total and no way to attribute them. So: no evidence it matters, which is not the same as evidence it doesn't.

The thing I would watch instead is issuance latency. That I have seen slow down on the high-volume account, from about two minutes to sometimes twenty, and it correlates with how many I have issued that month rather than in total.

A question nobody else will ask: over four years and $640,000, what does your bookkeeping look like? Not the tax position — simply the mechanics of turning a set of card balances with no statements into something you could reconcile if you ever needed to.

What's your monitoring? IME the failure that costs money isn't the decline, it's finding out about the decline four days later. Do you have anything watching the balances or is it manual?

Manual, embarrassingly, for the first three years. Now it is a script that logs into each dashboard on a schedule and writes the balance to a file, plus an alert if any of them fails to load twice in a row.

The alert on the dashboard failing to load turned out to be more useful than the balance alerts. Two of the five providers that died gave me a dashboard that stopped responding several days before anything else was visibly wrong, and in one of those cases I got most of a balance out because of it.

I'd share the script but it is four years of accumulated selectors held together by hope and it would break the day you ran it.

A dashboard that stops loading before the company dies. That's genuinely the most useful thing I've read this year!