Beers at the bar: the phone-calculator ending

The ritual at the end of the bar

The waiter sets the bill down on the table, and the ritual begins. One of the friends pulls out their phone, opens the calculator, the others stare at the bar's ceiling as if the answer were written up there. Someone asks whether the third round was paid by the one at the end or the one at the far side. Another remembers that the one with the glasses had only two beers because they're running tomorrow. One plate of bravas was shared by three, another by two, the third nobody explicitly ordered and it just appeared. Fifteen minutes later, the group pays, someone puts it on a card and the others send transfers for assorted amounts that don't tally to the cent, and there's always a euro missing or left over that just dissolves.

That scene repeats every Friday in thousands of bars. It's harmless, until it isn't. When the group has a long history of settlements like this, the imbalance piles up. There's always someone who suspects the group overcharges them, another who suspects they pay too little without meaning to, and a third who has lost track because they stopped following it three years ago. The cheerfulness of the ritual disguises a significant operational inefficiency: settling the bill for a group of six people shouldn't take fifteen minutes. It should take thirty seconds, and it can.

Why the bar bill is a badly modelled problem

The problem, stated well, is trivially simple. There's a set of items, each item has a price and a set of people who consumed it, and we want to calculate how much each person owes the payer. It's primary-school algebra. The reason it goes wrong at the bar isn't that the problem is hard. It's that it's modelled in the worst possible way: from human memory, at eleven at night, after three beers, with no written substrate.

Human memory, in a normal state, is mediocre with lists. In a relaxed, conversational state, it's downright bad. Add to this that no friend in the group is going to be willing to act as notary noting down each item with its consumer (and if you try, they'll call you a tightwad), and the "we'll split it at the end" model has two possible outcomes: the arithmetically equal split among everyone (unfair when consumption is very heterogeneous) or the by-eye split (systematically unfair to whoever consumes least and is most polite about paying).

Three settlement models and what they cost in time and resentment

1. Equal split without thinking

The cheapest model in time. The total bill is divided by the number of people and everyone pays the same. If the table is seven people, a bill of thirty-seven euros fifty, it comes to five thirty-six each. Done. Advantage: the settlement takes thirty seconds. Drawback: it penalises whoever ordered the alcohol-free beer or whoever left before the third round. If the group is stable and everyone consumes roughly the same over the year, this model works, because the deviations cancel out across nights. If the group is heterogeneous, it doesn't.

2. Split by actual consumption

The fairest model. Each pays for what they consumed, and the shared items are split among the actual consumers. Beer from the far end, two twenty euros a beer, two beers for one comes to four forty. Bravas eaten by three, nine euros among three is three euros a head. And so on. The classic objection is that this is slow and kills the mood at settlement time. The objection only holds if you model the problem badly, that is, you try to reconstruct it from human memory at the end of the night. Modelled well, it isn't slow: it's thirty seconds of tapping the items into an app and letting the split calculate itself.

3. The rotating round

The informal model. Each round is paid by one person, and over the nights it balances out because everyone pays equivalent rounds. It works reasonably well in small groups (four people at most) that are stable and see each other regularly. It fails in large groups, in groups where someone comes once a month and the others three times, and as soon as something destabilises it (someone leaves early, someone doesn't drink alcohol and the round is just beers). The rotating round isn't really a splitting system: it's a social heuristic that works when it works and breaks without warning.

The problem of shared dishes

The bravas, the croquettes, the plate of ham, the portion of squid. This is where the group loses the most time, because there are items ordered for the whole table (no friction: split equally among everyone) and items ordered by half the table but eaten by three (high friction, you have to remember who ate what). The trap is trying to solve it from memory. If instead it's noted in the moment, in an app, the item with the subset of people who consumed it, the problem vanishes. Noting three plates of bravas with three diners takes five seconds. Arguing over three plates of bravas at half eleven at night takes three minutes.

One additional important detail: portions split between two not as full consumers, but with a third who only "had a little", are the silliest source of friction at settlement time. The sensible engineering rule is binary: either you consumed it or you didn't, there's no "had a little". If the group applies this rule with discipline (and a little generosity towards whoever wants to taste), the heuristic works and the settlement speeds up. If the group opens the door to half-measures, the friction returns. Here the social convention the group gives itself matters more than the technical tool.

The efficient settlement: 30 seconds well spent

A well-done group settlement has three steps. One: someone writes down the items as they arrive at the table, marking the consumers. Two: when the bill is requested, the paper is checked against the receipt. Three: the system calculates and sends the transfers. The first step is the only one that requires discipline, and only if one person in the group takes it on (the most obsessive one in the group usually volunteers, delighted). The other two are automatic.

What makes this flow possible is having a tool that doesn't force you to do maths. Any decent system for this should let you record items with a subset of consumers, tolerate partially shared items, automatically calculate the split and split the cents by largest remainder, not by truncation, so the final payer doesn't always end up losing three cents of rounding. ControlarGastos does exactly that, and it lets you send the summary to the group in a message. The difference between the fifteen-minute ritual and the thirty-second settlement isn't the phone calculator. It's having modelled the problem in the moment instead of trying to reconstruct it at the end.

Conclusion: the friction at settlement is voluntary

The bar is a place for talking, not for doing arithmetic. Every minute the group spends squaring the bill is a minute it doesn't spend on the conversation you're all there for. That groups keep doing it by hand, with the phone calculator and with a fogged memory, isn't because it's inevitable. It's because nobody has raised their hand to say the way of doing it can be improved. Reducing the settlement to thirty seconds isn't geeky, it's civilised. And, above all, it frees the group from the one conversation no friend should ever have to have at the end of a good night: who owes whom and how much.