Why account friction, not math, decides who splits the bill

What actually decides whether a bill gets split?

Not the math. The math has been a solved problem for years — a calculator, a napkin, or now your phone’s camera can divide a total in seconds. What decides the outcome is how many people can get into the split without doing setup work. Every person you ask to download an app, create an account, verify a phone number, or remember a password is a person who might silently drop out before they ever pay. The bill doesn’t fail because the numbers were hard. It fails because the group never fully assembled inside whatever tool was supposed to hold it.

This is a different problem from the one most people name. Once everyone is in, there’s a second wall — agreeing on who had what, and paying across whatever apps people use — which is a coordination problem no calculator can solve. But that wall comes after this one. Pre-entry friction is upstream of all of it: you can’t coordinate, or even argue about fairness, with someone who never made it through the signup screen. The participation gate is the first filter, and it’s the one almost nobody designs for.

Why does forced account creation kill participation?

The clearest evidence comes from the place where the cost of friction is measured in lost money: e-commerce checkout. Baymard Institute, which maintains the most-cited dataset on why people abandon online purchases, finds that 18% of US online shoppers have abandoned an order because “the site wanted me to create an account.” (The list is a living one; it read 19% when this article was first published.) That’s not friction in the abstract. That’s nearly one in five shoppers who have walked away from an order because of a signup form.

Bill splitting is more fragile than checkout in one specific way. A shopper who abandons a cart costs the store one sale. But a split needs everyone. If one person bails at the signup wall, the other payments still arrive, but the settlement isn’t finished: the organizer is left with one share to chase for a week or quietly eat. The same checkout friction that costs a store a fraction of its orders leaves a dinner unsettled, because the organizer needs every participant, not most of them.

Source: Baymard Institute, Cart & Checkout Abandonment Rate Statistics, 2026 (living list; figures as of September 2026)

And the cost compounds with group size. Ask two people to install an app and maybe both do it. Ask six, and you need all six to clear the wall — each added person is another person who can stall. The friction isn’t paid once; it’s paid by every person you need, and you need all of them.

The login wall is two touches from “never mind”

User-experience research has studied this exact barrier for years. Nielsen Norman Group — the usability firm whose findings shape how serious apps are built — calls the demand to register before you can use something a “login wall,” and their verdict is blunt: it carries a “significant interaction cost,” so sites should use one only where users benefit significantly from it, with highly personal apps like email or banking as the examples.

“Users are utterly vexed to have to enter information before they get a taste of what is in store for them… People have to be highly motivated in order to incur that cost over the tempting alternative of deleting the app — which is pretty much two touches away.” — Raluca Budiu, Nielsen Norman Group, 2014

Read that against a bill split. A one-off dinner is the opposite of a banking app: low personal stakes, used rarely, and the “value” to the new user is paying money out. There is almost no motivation to climb the wall, and a frictionless escape — “I’ll Venmo you later” — sitting right there. Of all the contexts in which to demand account creation, asking a casual acquaintance to enroll just to hand back a few dollars is close to the worst. The person isn’t refusing to pay. The interface gave them an easier option than paying, and they took it.

Source: Raluca Budiu, Login Walls Stop Users in Their Tracks, Nielsen Norman Group, 2014

Friction beats motivation: what the behavior model says

There’s a tidy framework that explains why “they just didn’t want to pay” is usually the wrong diagnosis. The Fogg Behavior Model, developed by Stanford’s B.J. Fogg, holds that a behavior occurs only when three elements converge at once: motivation, ability, and a prompt (B = MAP). If any one is missing, the behavior doesn’t happen — and crucially, motivation and ability compensate for each other. The harder something is to do, the more motivation it takes to do it anyway.

A signup wall doesn’t need to touch motivation to stop the behavior. Grant that the person still wants to settle up; the wall attacks ability, which the model breaks into six factors, among them time, money, physical effort and “brain cycles.” Creating an account spends three of them at once: time (minutes you don’t have), physical effort (thumbs on a tiny keyboard) and brain cycles (a password to invent and remember, a verification code from a text). On the model’s terms each one lowers ability, and stack enough of them and the behavior drops below what the model calls the action line — the threshold at which behavior occurs. That is the model’s prediction, not a measurement of dinner guests; but the fix it points to isn’t more nagging. It’s raising ability by removing steps.

Sources: B.J. Fogg, A Behavior Model for Persuasive Design, Proceedings of the 4th International Conference on Persuasive Technology, 2009 (the model’s origin); the B = MAP formulation, ability factors and action line as summarized in Fogg Behavior Model, The Decision Lab

Who actually pays for the friction

The cost of a signup wall doesn’t land on the app. It lands on the organizer — the person who fronted the bill and now has to collect. Every guest who stalls at the wall becomes a payment the organizer has to chase by text, or write off. The friction you can’t see on the screen shows up later as a number in someone’s pocket.

It is tempting to turn Baymard’s 18% into a per-person stall rate for a dinner table. Don’t. It is the share of US online shoppers who say they have ever abandoned an order for that reason, not the odds that any given guest balks tonight, and nobody has measured how often a dinner guest gives up at a signup screen. What survives the translation is the structure, not a number: the organizer needs every guest to clear the wall, and each additional guest is one more person who has to clear it. A bigger table doesn’t just mean more math. It means more walls to clear, and more chances the settlement comes back with a hole in it.

Economist Mancur Olson spent a career on a version of this problem. His 1965 book on collective action argued that group efforts run into a free-rider problem — many of the people who would gain from the effort get the benefits even if they bear none of the costs — and, as the Econlib summary of his work puts it, his argument was not that collective action fails but that those who organize it need to find some way of giving people incentives to join in paying for it. Olson wrote about large groups and lobbies, not dinner tables; the application is ours. A bill split is a tiny collective effort that needs every member to pay in, and the cheapest incentive an organizer has is to make paying in cost almost nothing. Hand the group a signup wall and you’ve done the opposite, to the very people whose participation you can’t do without.

Source: Mancur Olson, The Logic of Collective Action (Harvard University Press, 1965), as summarized in Econlib’s biography of Olson

That asymmetry is the case against treating a dinner like a checkout. A store optimizes for the average cart; an organizer needs a perfect collection. The only reliable way to get one is to stop asking the people you need to pay to do setup work they have every reason to skip.

“But isn’t an account good for trust and records?”

It’s a fair objection, and the honest answer is: sometimes. An account is genuinely worth the friction when the relationship is ongoing and the stakes are high — which is exactly the case Nielsen Norman Group carves out for login walls. A long-term roommate ledger that tracks balances for years, like Splitwise, has a real reason to ask its members to sign in; that’s its whole model, and the persistence is the point.

The mistake is applying the high-friction pattern to the low-stakes job. A one-time dinner doesn’t need everyone to have an identity in your system. It needs everyone to pay once and leave. The trust and record-keeping that an account provides are real benefits — for whoever maintains the record. Forcing them onto every guest is paying the cost without getting the benefit.

What does an account actually buy a shared ledger? An author.

Every entry in a record that other people will be held to needs someone who can be held to it, and an account is how the ledger knows who that is. That is a need of the writer, not the payer — and the two have been blurred together for so long that the wall gets built in front of everyone.

Paying your share of tonight’s check is a read followed by a payment. You look at a number, you agree it’s yours, you send it from an app you already have. Nothing about that operation changes anyone else’s number. Adding an expense to a running roommate ledger is a write: the figure you type becomes what three other people owe next month. Those are different operations with different requirements, and only one of them needs to know who you are.

Accounting theory has a name for what the write operation needs. Yuji Ijiri put accountability at the center of his account of what accounting is: in his accountability-based framework, what separates accounting from every other information system is that it runs between two parties: an accountor, who is obliged to account, and an accountee, who is owed the account. The objective of an accounting system, in Ijiri’s words, is “to provide a fair system of information flow between the accountor and the accountee.” In 1975 he had put the point more bluntly: “accountability is what distinguishes accounting from other information systems in an organization or in a society.” The relationship comes first. A record that cannot say who is answering to whom is information, but it is not an account. The parallel to a shared expense ledger is ours, not Ijiri’s; his subject was accounting systems, and the point we take from him is narrower: a record built to hold people to it needs a named party on each side.

Sources: Yuji Ijiri, On the Accountability-Based Conceptual Framework of Accounting, Journal of Accounting and Public Policy, 1983. Both lines are quoted as reproduced (Ijiri 1975, p. 32; Ijiri 1983, p. 80) by Tae Wan Kim, Pierre Jinghong Liang and John Hooker in their February 2021 Tepper School working paper, published in 2024 as Yuji Ijiri’s Fairness Question in Accounting, Accounting, Economics, and Law: A Convivium

This is why the account-based ledgers ask for the wall, and it is worth reading their own reason rather than assuming one. Splitwise’s help center explains why it wants members invited by email or phone, and the reason is a records policy, not a growth tactic:

“We think it’s important that everyone in a Splitwise group can log in and see their balance. When you invite friends using their email address or phone number, they can view expenses, fix mistakes, and add their own, which helps avoid disagreements down the road.” — Splitwise Help Center, “Can I add a friend without adding their email address or phone number?”

Read that through Ijiri’s frame. Each member of the group is both accountor and accountee, and the login is what binds a person to the entries they write and lets them contest the ones they didn’t. The same page allows a placeholder friend, “someone tracked by name, without contact info,” and notes that you “can always add their email or phone number later to give them access to your group,” and it lets you keep adding expenses with an invited friend even if they never accept. So the wall is not in front of being counted. It is in front of logging in, seeing your balance, fixing a mistake or adding an entry — the operations where a member reads and writes the record as themselves. The wall is the price of attribution, and it lands where the record is touched, not where the bill is paid. That is a fair price for a ledger you will live inside for a year. It is the wrong price for a check.

Source: Can I add a friend without adding their email address or phone number?, Splitwise Help Center

Why “anyone with the link can add expenses” is the wrong fix

Because it keeps the ledger and throws away the author. A newer wave of Splitwise alternatives sells exactly this trade. One markets itself on the promise that “anyone with the link can add expenses and view balances — no account, email, or password,” and calls the group private because “only people with the link can see the group.” That removes the wall this article is about. It removes it for the wrong operation.

Source: The Free Splitwise Alternative — No Signup Required, Bill Splitting App, 2026

The web’s standards body has a name for a link that works like this: a capability URL. In the W3C Technical Architecture Group’s 2014 draft on the pattern, “capability URLs grant access to a resource to anyone who has the URL,” and the draft’s central warning is that “URLs cannot generally be kept secret.” They sit in browser history and server logs. They can leak through the Referer header when someone clicks out to another site. They are exposed to whatever service the link was shared through. The draft’s advice to anyone issuing such a link is that the page “should describe what people who get the URL can do with it.” For a read-only page, what they can do is look. For a ledger with link-write access, what they can do is change what you owe.

Source: Jeni Tennison (ed.), Good Practices for Capability URLs, W3C Technical Architecture Group, First Public Working Draft, February 2014

The draft does not ban the pattern. It names cases where a capability URL is useful and asks issuers to say what the link can do. What follows is our reading for a shared ledger, not the W3C’s. When the URL is the credential, every entry has the same author: whoever had the link — a name typed on an unauthenticated page is a label, not an author, because the record cannot verify it. That is a tolerable trade when the link’s only job is to show a share and take a payment: the exposure is a read, someone who shouldn’t have the link seeing one share of a restaurant bill, which is not nothing but changes no one’s number. It is a different thing when the link is a running record. The forwarded group-chat link, the roommate’s ex, the old phone — each one can write, and the ledger cannot prove which of them, or you, did. On our reading of Ijiri’s frame, the record then has no accountor it can name, and his “fair system of information flow” collapses into a shared notepad with a balance at the bottom.

Elinor Ostrom found the same requirement in a very different place. Her Nobel lecture reproduces, in the updated wording of Cox, Arnold and Villamayor-Tomás, the design principles she drew from common-pool resource regimes that survived for long periods, and principle 4A is monitoring users: “Individuals who are accountable to or are the users monitor the appropriation and provision levels of the users.” A shared household ledger is a very small commons — a pool everyone draws on and everyone is supposed to top up. Ostrom’s principle is about resource regimes, not apps, and it says nothing about ledgers or accounts; the extension is ours. But monitoring, on any reading, gets much harder when no draw has a name the record can vouch for, and a ledger that anyone with the link can write to leaves no way to verify who took what.

Source: Elinor Ostrom, Beyond Markets and States: Polycentric Governance of Complex Economic Systems, Nobel Prize Lecture, December 2009; published in American Economic Review, 2010

The design that follows is neither “accounts for everyone” nor “a link for everyone.” It is to put the writing in one place. In splitty, the person who scanned the receipt is the one who assigns items; everyone else opens a Share Split link and pays their share with their own preferred method, without the app. The link carries one person’s share and a way to pay it — not a ledger that everyone holding the URL keeps adding to. And because splitty settles one bill now rather than keeping a running balance over time, there is no long-lived record for a stray link to quietly rewrite. The organizer does the setup. The guests just pay.

What zero-friction entry actually looks like

The design principle that follows is simple: concentrate the setup on one person, and ask nothing of everyone else. The organizer — the one motivated enough to herd the bill — can install a tool and do the work. The other five should meet the split the way they’d meet a text message: it’s just there, it shows their share, and paying happens in an app they already trust. No download, no account, no wall.

E-commerce learned this lesson the expensive way and then fixed it. Nielsen Norman Group’s guidance on registration is blunt about the payoff: “Forcing registration causes lost sales… It is common for sites that add guest checkout to immediately realize increased sales.” Guest checkout is the same move — let the person complete the task first, and make the account optional, for the one user who actually wants it. A bill split should work the same way: the guest’s job is to pay, not to enroll.

Source: Amy Schade, Don’t Force Users to Register Before They Can Buy, Nielsen Norman Group, 2015

This is the model a shared payment link gestures at, and the one splitty is built around. One person scans the receipt and assigns items; everyone else gets a Share Split link they open in a browser, see exactly what they owe, and pay with their own preferred method — Venmo, Cash App and PayPal among them. There’s no account to make and nothing to download for the people being asked to pay. The friction that stalls people at the wall is simply removed from their path, because the wall was never built for them in the first place.

Once everyone’s actually in, the next problem is the one we cover elsewhere: getting a mixed group to agree on the split and settle it across different apps. But that’s a problem you only get to have if everyone made it through the door. Solve entry first. It’s the cheapest fix in the whole chain — and the one almost everyone skips straight past on their way to optimizing the math.

Enrollment is also the clearest way to tell the approaches apart. Compare the three moments at which a bill can be divided: a shared card that splits at the swipe needs everyone signed up before the meal, while a receipt-side split needs exactly one person to have anything installed. That is the same wall, priced differently.