Average order is $4.20 and I do ~900 a day — every gateway I try eats the margin

Everyone here is optimising for the four hundred dollar invoice and I have the opposite problem.

Average order value is $4.20. I do somewhere between 800 and 1000 of them a day. Margin on each one is roughly 90 cents before payments. Every gateway I've tried quotes me a percentage that sounds trivial and then the actual settlement lands short by an amount that has nothing to do with the percentage.

I know why. I want to know what people at this shape of volume actually run, because I cannot be the only one.

You already know why but let's put the numbers on it, because the fix depends on which of the two costs is biting.

At $4.20 a percentage fee is not your problem. Coinremitter at 0.23% takes a cent. NowPayments at 0.5% takes two. Neither of those is visible against a 90 cent margin.

The killer is anything fixed. Here is what a single $4.20 order costs on each rail, using what I paid in the last week of February:

Rail	% fee	Fixed cost	Total	% of order
BTC on-chain	0.01	0.41	0.42	10.0%
ETH L1	0.02	0.94	0.96	22.9%
USDT on Tron	0.02	0.12	0.14	3.3%
BTC Lightning	0.01	0.00	0.01	0.2%

On-chain Bitcoin at your ticket size is not a payment method, it's a donation. Whatever gateway you pick, the decision that matters is which chain the customer pays on, and most of them will pay on whatever the checkout shows first. Reorder the list and half your problem goes away before you change providers at all.

Caveat on my own numbers: one week, and February was a quiet fee week on L1. In a busy one the ETH row gets worse, not better.

stop letting them pay in btc onchain. that's the whole thread

On tron the fixed cost is a rounding error and it has been for years. What are your buyers actually holding that makes this hard?

Mostly BTC. It's a technical audience and they're stubborn about it. When I've defaulted the checkout to a stablecoin I get support mail about it.

If they insist on Bitcoin then Lightning is the only version of that which works at $4.20, and you've got two doors.

OpenNode is the hosted one — 1% flat, Lightning-first, on-chain fallback. That 1% is four cents on your order and it's the whole cost. The catch is the signup: they ask for business information, they're a US operator, and if you want to settle in USD rather than BTC that triggers full KYC. So it's the fast option, not the quiet one.

BTCPayServer with your own LN node is the quiet one and costs zero percent. You already run servers, so you know what that sentence is hiding.

Before anyone recommends running your own node for this — 900 inbound Lightning payments a day means channel liquidity is now a daily job. You will run out of inbound capacity mid-afternoon and every invoice will fail while you're asleep. That's not a fee problem, it's a rebalancing problem, and it doesn't go away by paying zero percent.

This is the objection I came in already worried about, and it's the correct one.

# what I actually want alerting on, and neither dashboard gives me
lncli listchannels | jq '[.channels[] | .remote_balance|tonumber] | add'
# inbound capacity across all channels. below one day of turnover = page me

Hosted at 1% is four cents. Me being woken up is worth more than four cents.

What is the product at $4.20 times 900? That number is doing a lot of work in this thread and nobody's asked.

Different idea. Stop charge per order.

Sell credit pack. 20 dollar, 40 dollar, whatever. Customer pay one time, you take one payment instead of five, fee become nothing per action. This is what everybody selling small thing figure out eventually, is not new.

Anyway then your problem is refund of unused credit, which is smaller problem than 900 invoice a day. Also you get money before you deliver, which is not bad.

Annoyed that this is right. I've been solving the payment problem when the pricing model is what's broken.

Not changing it this quarter — the whole product is built around per-action billing and unpicking that is a month. But I'm not going to pretend a gateway choice fixes it either. Leaving this unsolved on purpose.