Split at the swipe vs. split at the receipt: what each sees

When does a split actually happen?

At one of three moments. Naming them is most of the work, because once you have the names you can see which one any given product has picked, and what it gave up to pick it.

One clarification, because it is the thing most easily muddled: this is a taxonomy of when the division is computed, not of when the money finally lands. The two can sit at different moments, and often do — splitty computes at the receipt and collects at the request, which is why it inherits the accuracy of the third moment and the collection risk of the second. The swipe is the only one where computing and collecting are the same event.

Most published comparisons of bill-splitting apps sort them by feature list. Sorting them by moment is more useful, because the moment of computation is what constrains the feature list. A product computing the division at the swipe cannot itemize from the transaction, because a card authorization carries a total and not a list. It could of course go and fetch the receipt — but that is the third moment, not the first, and it inherits the third moment’s timing.

Moment one: the swipe

The swipe-side pitch is genuinely strong, and it solves the problem the other two moments cannot. If everyone’s card is charged at the moment of payment, there is no fronter, no ledger, and no awkward follow-up. The collection problem is not managed — at the moment of payment, it does not arise.

Cino, a shared-card product aimed at couples and housemates, describes the mechanism plainly: you “link your personal debit cards — you don’t need to open a new joint account,” you “set your custom split ratio — 50/50, 60/40, 70/30, or anything you agree on,” and then “every payment is split automatically between you both in real time.” No shared bank account, no top-ups, and both people keep their own bank.

Source: The easiest way to split bills based on income, Cino, November 2025

The number at the swipe is not always the final number

Here is the part that rarely makes it into a product page. A card authorization and a card settlement can be two different numbers, and the payments industry has built formal machinery around the gap between them.

Visa’s own guidance is explicit about why estimated authorizations exist: they are “used by merchants in scenarios and segments where the final amount is not known.” When the estimate turns out to be too high, “merchants must process partial reversals for the difference.” When it is too low, the merchant “must use an incremental authorization or a second authorization for the additional amount.”

The scope is worth stating precisely, because it is narrower than you might guess. Visa’s 2019 expansion opened estimated and incremental authorizations to parking (MCC 7523), electric vehicle charging (MCC 5552) and card-absent grocery (MCC 5411), and using them at all is “optional for merchants.” No restaurant category is on that list. What the rule demonstrates is not that every terminal shows a placeholder — it is that the networks consider “the final amount is not known” a common enough situation to write settlement machinery for it.

Restaurants have their own version of the gap, and MasterCard’s rule is written for the case where the card is not physically present. As the payment-processor documentation describes it, in 2017 MasterCard reinstated a 20% tip tolerance for card-not-present transactions conducted by US restaurant merchants under MCC 5812 and 5814 — a rule whose stated purpose is to enable “a restaurant merchant to add a tip amount after the authorization was approved.” The documented cases are a sushi order phoned in and collected at the counter, and a pizza delivered to the door: the receipt prints with a tip line, the guest writes in an amount, and that amount lands after approval. The tip must stay within 20% of the authorized amount; the portion beyond that “may be charged back by the issuer for no authorization.”

This is not a knock on any one company. It is a property of the moment. You cannot divide a total that has not been finalized, and at the swipe there is no guarantee that it has been.

Sources: Expanded Eligibility for Estimates and Incremental Authorizations (AI09108), Visa; MasterCard: 20% Tip Tolerance Reinstated for Card-not-present Restaurants, Toast

A ratio agreed before the check existed

The second constraint is sharper. A shared card has to know how to divide before it knows what it is dividing, so it divides by a rule set in advance: 50/50, 60/40, an income ratio. As Cino puts it, “once your split ratio is set, every transaction is divided instantly between your accounts.” The rule is not frozen — Cino supports exact amounts as well as percentages, and you can change the ratio whenever you like. What you cannot do is set it in response to a check you have not seen yet.

For rent, that is exactly right. Rent is a single indivisible number that two people consume jointly, which is why an income-proportional system works well for household bills. A restaurant check is the opposite kind of object. It is a list of things individual people ordered.

splitty’s own scanned receipts make the point. Across splitty’s US-leaning restaurant receipts, nine of the ten most-frequent line items are single-serve drinks — diet coke, espresso martini, water, coke, coffee, Tito’s, aperol spritz, margarita, soda. The only non-drink in the top ten is french fries. The items that recur most often across those receipts are not shared plates. They are things one person orders for themselves.

That is the whole argument in one line. Tap water and an espresso martini sit in the same top ten, and a ratio agreed in the abstract cannot tell them apart. It was never given the chance — it was set before anyone ordered.

Source: splitty first-party data — most-frequent line items across splitty’s US-leaning scanned restaurant receipts, July 2026 snapshot. Item frequency only; splitty’s snapshot does not record item prices, and these shares describe splitty’s own users rather than US diners generally.

The salience cost of paying automatically

There is a third cost, and it is the least obvious. Automatic payment does not just move money without friction — it removes the occasion on which anyone has to look at the charge.

The cleanest measurement of this comes from utilities. Steven Sexton, writing in The Review of Economics and Statistics in 2015, studied what happened when customers enrolled in automatic bill payment programs that “eliminated the need for consumers to view recurring bills.” His argument: if enrollees stop inspecting their bills, price salience declines, the price a boundedly rational person perceives falls, and consumption rises.

It does. Enrollment in automatic bill payment increased residential electricity consumption by 4.0% and commercial electricity consumption by as much as 8.1%. Programs that smoothed seasonal variation in low-income customers’ monthly bills produced 6.7% greater electricity use.

Electricity is not dinner, and the disanalogy is worth stating rather than glossing: autopay acts on a stream of ongoing consumption decisions, while a restaurant split lands after everyone has already ordered. Nothing here says a shared card raises restaurant spending by 4%, and the ordering is already done by the time the ratio applies. What Sexton documents is the consumption effect; the channel he proposes for it is that enrollees stop inspecting their bills. He measures the electricity, not the attention — so read it as the reason to expect that a split which resolves itself silently is a split nobody re-reads, not as proof of it. The line items that most often surprise people on a restaurant bill are exactly the ones an automatic division would sail past.

Source: Automatic Bill Payment and Salience Effects: Evidence from Electricity Consumption, Steven Sexton, The Review of Economics and Statistics 97(2), 2015

Moment two: the request

The middle moment is the one almost everyone actually uses. One card pays, and the payer sends requests afterward. Its advantage over the swipe is that the amounts are free — they can reflect anything, including a careful itemized reading, because they are computed after the total is known.

Its cost is equally well known, and this site has covered it at length rather than re-deriving it here: the request is a promise, not a payment, which is why “I’ll Venmo you later” so often never happens, and why the signup wall, not the arithmetic, is what usually kills a split. The request moment converts a payments problem into a social one.

Worth noting what the request moment shares with the swipe: both can be run on a number nobody checked. Sending an even-split request for a check you never read is the same failure as letting a ratio divide it silently — it just fails politely.

Moment three: the receipt

The last moment has the most information and the least leverage. The itemized receipt is the one artifact that lists what was actually ordered, and the copy you scan after settling carries the tip and any service charge printed on it. That is a claim about the finished receipt, not about every piece of paper a restaurant hands you — the tip-line receipt in MasterCard’s own example is printed precisely so a tip can be written on it afterwards. The catch is constant either way: by the time the document is complete, the money is already gone from somebody’s account.

Restaurants can do this at the terminal, and the mechanics are worth knowing because they show the ceiling. Toast’s point-of-sale lets a server split a check by item — tap items and move them onto a new check, or split a single item any number of ways. Two limits are instructive. The dedicated split button works up to nine ways, and beyond that a server has to fall back to manual payment entry — with one geographic exception, since Toast documents an enhanced “Split evenly” route past the cap for customers in Canada, Ireland and the U.K. only — one of several vendor settings that decide whether a table can split at all. And when a check is split by item, “you will have to add a service charge on each split check manually” — the service charge does not follow the items.

That second limit matters more than it looks. Splitting by item at the POS divides the food and then makes a human re-derive the fees, which is precisely the step people get wrong. A mandatory service-charge or auto-gratuity line shows up on 1.8% of splitty’s scanned US restaurant receipts — read that as a presence check rather than a trend, and as a floor at that, since it only catches charges printed as their own line. Uncommon, in other words. Rare steps are the easy ones to forget — though that is a judgment about human routine, not something these receipts measure.

The POS route also has a timing gate that has nothing to do with software. Toast’s own instructions note that splitting before the order goes to the kitchen prints separate tickets, while doing it afterwards is purely a payment-side operation. In practice the ask gets harder the longer you leave it, which is the argument of our guide to when to ask for separate checks.

Sources: Split Checks on the POS, Toast; splitty first-party data — service-charge prevalence across splitty’s US-leaning scanned restaurant receipts, July 2026 snapshot

The trade-off nobody advertises

Lined up side by side, the three moments trade the same two things against each other. Read the table as how each moment is typically implemented rather than as a law of nature: a terminal multi-card payment need not use a preset ratio, and not every receipt-side tool allocates tax proportionally. What does not vary is the row that matters — what information exists at the instant the division is computed.

Read the last two rows together and the shape of the trade is obvious. The swipe buys certainty of collection with everyone’s enrollment and all of the detail. The receipt buys all of the detail with the risk that someone shrugs.

Where splitty sits, and what it does not fix

splitty is a receipt-side product, which means it takes the receipt-side deal in full — the good half and the bad half.

The good half: it works on the finished check. Scan the itemized receipt and every item starts split among everyone; you tap to remove the people who didn’t have it, and tax and tip distribute proportionally to what each person actually ordered. Because it reads the printed receipt, the service charge and the tip are already in the number being divided, not a true-up waiting to happen. Only one person needs the app — everyone else receives a pre-filled request, or a web link they can open without downloading anything.

The bad half, stated plainly: splitty does not eliminate the pay-later gap. It sends a request, and a request is still an ask. A shared card genuinely beats it on that specific dimension, and no amount of receipt accuracy closes the difference. What splitty argues is that the ask is easier to win when the number attached to it is visibly correct and itemized — not that the ask disappears.

Which moment should you use?

The answer is not the same for every group, and it follows almost entirely from two questions: does everyone consume roughly the same thing, and do you trust everyone to pay?

One mechanic to avoid regardless of moment: letting chance decide. Some shared cards ship a feature that picks a random payer for the night, which is a cheerful version of the credit card roulette trap — entertaining once, and a slow transfer of money from unlucky people to lucky ones over a year of dinners.