Notes from Nic Carter's Transaction Count is an Inferior Measure: https://medium.com/@nic__carter/transaction-count-is-an-inferior-measure-fba2d5ac97f1
This article nicely brings to the surface a few important and uncomplicated details about our present payment systems like Visa, Zelle, and Bitcoin. It helped me to contextualize the role of the different implementations of payment rails and how we discuss them.
1. In crypto, individual projects will use transactions per second (tps) as a means to compare themselves to other projects and to the large established systems like credit card payment processors. Unsurprisingly, this single variable isn't very informative.
2. To more accurately capture the intended behavior or utility of payment networks, we should describe them in terms of their economic throughput ( [average transaction size] * [number of transactions] ) and their settlement assurances, a neat quantitative/ qualitative pair.
3. The author graphs several real world payment systems and demonstrates that as throughput increases, payment systems tend to aggregate into payment networks transmissing small balances at high tps (e.g., cash, credit cards, Venmo) and settlement networks transmitting large balances at low tps (e.g., FedWire, Bitcoin, gold).
4. In terms of throughput alone, I was surprised to read that Bitcoin is only one order of magnitude away from Visa, which I was very surprised (and pleased) by. The implementation of the settlement assurances is where the trade offs occur that separate the two groups.
5. Credit card companies leverage fast payment systems for small everyday balances, but are centralized (permissioned), require already trusted rails (e.g., banks), and have long settlement times.
6. Better said: "Implementing long settlement times allows for fast payments b/c when trusted parties interact, error correction and ledger alterations after the fact (chargebacks) are possible and desirable."
7. A settlement network like Bitcoin or FedWire performs less transactions per second, but offers rapid, more permanent, and usually more costly settlement assurances. Transactions occur less frequently but when they do, they are "crystallized" immediately.
8. The author calls this "money for enemies," where once a transaction is completed, there is no recourse, as little surface area possible for cheating or reversal after the fact.
9. So it's clear that the single variable "tps" is too non-specific to compare payment networks as it doesn't capture the determining factors that make any payment network useful.
10. We can go a step further and say that payment networks also can't be meaningfully compared to each other except in the context of a specific use case. The best network to use in any particular scenario is that which best satisfies the needs of the transacting agents.