Company
Card pricing is not confusing by accident. The distance between the rate a business is quoted and the rate it actually pays is worth money to whoever is on the other side of the contract, so nobody on that side is in a hurry to close it. Pyncht exists to close it, and to show its working while it does.
See the sample analysisThe thesis
What you are charged, what you should be charged, and what you never collected at all.
A card payment is never one fee. It is interchange, set by the bank that issued the card and rewritten a few times a year; a scheme fee, set by Visa or Mastercard; and a markup, set by your processor. Interchange-plus pricing itemises all three, which is why it looks complicated. Blended pricing shows you a single number — 2.9% and 30¢ — and quietly keeps the difference between what the networks charged and what you were billed. Neither model is dishonest. Only one of them is legible.
So the quoted rate is a headline, not a price. The effective rate — total fees divided by total volume — is what actually happened, and it moves with things no contract mentions: your card mix, where a corporate card can carry 2.45% of interchange while a debit card carries under 0.85%; the share of volume that crosses a border; your average ticket measured against a fixed per-transaction fee. Two businesses on identical paper contracts routinely pay rates thirty or forty basis points apart.
Fees are the line every finance team looks at. Declines are the larger one, and they appear on no invoice at all. An authorisation rate of 91.88% means roughly one charge in twelve never completes, and those reason codes are not equally final: expired cards recover at 88% once an account updater refreshes the credential, issuer-unavailable at 94% on a retry an hour later, insufficient funds at 62% if you simply wait for payday. In the sample analysis the recoverable revenue is $96,561 a year against $13,901 of fee savings — seven times larger than the number everyone was arguing about, and invisible in a processing statement.
Put the three together and payments stop behaving like a fixed cost. An effective rate becomes something you can move: settle European volume locally, pass Level 2 data on commercial cards, retry only the declines that actually recover. Each change is individually small and individually boring. In the sample analysis they total $110,463 a year on a $3.2M run rate — about 3.4% of revenue that was never a product problem, a pricing problem, or a demand problem. Pyncht is the layer that finds those changes and, when you decide it should, makes them. It is not another processor.
Principles
Not values on a wall. Each one is a decision already made in the interface, and each one costs us something.
Every figure Pyncht reports carries its own calculation. An opportunity is never asserted at a dollar value; it shows the volume it applies to, the rate it is multiplied by, the network rule that sets those basis points, and the confidence we hold it at. A number you cannot reproduce is a number you cannot take into a processor negotiation, which makes it worthless at exactly the moment it matters.
Fees you avoid and sales you recover are different money with different certainty, and adding them together is the oldest trick in this category. Pyncht reports them in separate columns everywhere they appear. In the sample analysis that is $13,901 of fee savings and $96,561 of recovered revenue, totalling $110,463 — three figures, kept as three figures, because collapsing them into one would flatter us.
Connecting a processor grants Pyncht permission to read settlement data and nothing else. No charge is routed, retried, or repriced until you switch that specific change on, one at a time, with the projected impact and the reasoning shown before you commit to it. The analysis is designed to be worth running on its own — including by businesses that never move a cent of volume to us.
If a figure cannot be traced to a source record or a published network rule, it does not ship. That is why this page carries no customer count, no uptime percentage, no funding history and no logo wall: Pyncht is a working prototype, the analysis you can open is modelled data, and it says so on every screen it appears on.
By the numbers
Counts of what the cost model implements today. No customers, no headcount, no round — those numbers would be easy to write and impossible to defend.
Every count above is a property of the shipping cost engine, not a roadmap. The merchant it runs against is illustrative and named as such throughout.
Careers
Small team, unusually specific problems: modelling issuer behaviour, reconciling settlement files that disagree with each other, explaining a basis point to someone who has never had to care about one. Roles are listed only while they are genuinely open — there are no evergreen postings here.
Connect a processor read-only and Pyncht rebuilds twelve months of true cost, fee by fee. Nothing routes, retries, or reprices until you say so.