Does splitting the check remove your discount?

What actually happens when a check is split by item?

Mechanically, splitting by item is a move. The server opens a split screen, taps items, and drops them onto a new check. The items travel. The question is what travels with them — and the answer is that a line item is not a self-contained object. It carries a price, but the discount attached to it may be a separate record that points at the original check.

Toast is unusually direct about this in the support article its own restaurants read:

“Item-level discounts also do not carry over when you split a check — the discount is removed from the item once it moves to the new check, so you’ll need to reapply any discounts on each split check individually.”

Read that as a workflow instruction and it is unremarkable: staff are told to reapply discounts. Read it as a diner and it is a warning. The documented fact is narrow but sufficient — reapplication is manual, so the discount survives only if someone performs an extra step. How often that step gets missed is not something Toast publishes, and we are not going to pretend otherwise. The same document notes that service charges do not translate across split checks either — a related failure we cover in the piece on what a split sees at the swipe versus at the receipt, which digs into the terminal mechanics and the nine-way cap in depth.

Source: Split Checks on the POS, Toast

Square does it differently — and creates a different problem

It would be convenient if every point-of-sale behaved like Toast. They do not, and assuming they do is how you end up arguing with a server who is correctly telling you their system works another way.

Square’s support article on splitting a check states that when a check is split by seat, discounts, coupons and service charges that are applied to specific items, or configured to be item specific, move alongside the item they are associated with. So far, better news. But the same paragraph adds the catch:

“If there are any cart-level service charges, discounts, or promotions, then those would remain on the original check and can be moved manually.”

That is a different failure with a different shape. Nothing is documented as deleted here — the discount is stranded. A whole-table promotion stays attached to whichever check is the original, and the other checks are created without it. Square’s documentation stops there, so this page will too: it says where the discount stays, not what it is then worth or what the group’s combined total comes to. The part that matters at the table is simply that everyone on a newly created check is paying an undiscounted share unless someone moves the discount by hand.

Source: Split a check by item or seat, Square

The third case: the deal stops being true

There is a category of discount that does not get moved or stranded, but invalidated — because splitting the check destroys the condition the discount was testing.

Buy-one-get-one is the clean example. A BOGO is a rule about a pair of items on one check. Split those two items onto two checks and each check now holds a single item, so the rule no longer sees a qualifying pair. In a 2025 thread on Square’s seller community, a restaurant operator described exactly this: a weekly BOGO that “works quite naturally on the POS — until you split the items across seats,” after which the automatic discount stops applying.

The reply that the community marked as the solution — written by a Square Champion who states plainly that he is a fellow seller and not a Square employee, so treat it as a well-informed merchant account rather than vendor documentation — explains the mechanism: an even split only prorates the total and moves nothing, while splitting items across seats moves them into their own transactions, so “the automatic discount rules don’t recognize that the BOGO was satisfied.” A second operator replied months later insisting it was a regression rather than intended behaviour. Either way, the diner-facing result is the same.

What is actually established here is one vendor, one discount type, one merchant’s report. The reason it still matters is the mechanism rather than the sample: any conditional promotion — spend $50, get $10 off, buy two entrées and get an appetizer free, a prix-fixe requiring three courses — is a claim about a whole check, so a rule engine evaluating a smaller check could reach a different answer. Treat that as a question to ask rather than a documented outcome: nobody has published what each system does with each offer type, and we are not going to assert it for them.

Source: Automatic Discounts and Split Items Workaround, Square Seller Community, 2025 (merchant discussion, not official documentation)

Which discounts survive a split?

Ask one question before the check is divided: what is this discount attached to? The answer predicts the failure.

One caution about reading that table: each cell reports what a named vendor documents for a named operation, not a universal rule. Square’s note describes splitting by seat; Toast’s describes splitting by item. Other systems, and other split methods on these same systems, may behave differently and are simply not covered by either document.

That last row is the one people get backwards, and it is the exception that rewards attention: Toast’s documentation says that if guests want to earn loyalty points across multiple loyalty accounts on a single order, the check should be split rather than the payment — splitting the payment instead means points will not accrue properly. So for loyalty, asking for a genuine check split is the move that protects you. For item-level discounts on the same system, it is the move that endangers them.

Sources: Split Checks on the POS, Toast; Split a check by item or seat, Square

Is the discounted line really where pricing goes wrong?

Partly — and the research is more specific than the obvious version of this argument, in a way worth reporting honestly because some of it cuts the other way. The literature on checkout price accuracy does not find that discounts in general are error-prone. It finds that advertised specials are, while other kinds of price reduction are not.

The reference point is the FTC’s 1998 Price Check II — described by NIST as a larger, more comprehensive review than the 1996 survey that preceded it — run using the price-verification procedure NIST and the National Conference on Weights and Measures developed jointly, which 42 states had adopted. It compared scanned prices with the lowest posted or advertised price across 107,096 items. Overall accuracy was decent — one in 30 items was mispriced, and errors split roughly evenly between overcharges and undercharges. Sale items were the exception on both counts.

The FTC’s own consumer-protection director drew the practical conclusion at the time: consumers can protect themselves by paying close attention to the prices they are charged, particularly when buying sale items, because almost two-thirds of the scanner errors in that category were overcharges. The inspectors also catalogued the causes, and they are mundane and familiar — out-of-date signs, incorrect prices in the computer, shelf prices that never got updated. Several of those are failures to propagate a price change, which is a plausible reason advertised prices would be the fragile ones — though the inspectors catalogued causes rather than testing that explanation, so read it as a hypothesis, not a finding.

Sixteen years later, a team of pricing researchers went back to that same FTC dataset and re-ran it in the Journal of Retailing, this time measuring the size of each error rather than only its direction. The asymmetry held: undercharges ran about 22% for both sale and non-sale items, while average overcharges were 19% on non-sale items and 38% on sale items. The same 2014 paper sets those figures against the work that opened this line of research: it reports that Goodstein’s 1994 study in the Journal of Marketing found average overcharges of 6.9% against average undercharges of 23.7%. Worth flagging, in the same spirit as the caveats elsewhere on this page: Goodstein’s original 1994 article sits behind a publisher paywall we could not open directly, so those two figures reach this page as quoted inside the 2014 re-analysis, not independently re-verified against the 1994 text itself. On that basis, the 2014 authors’ reading is that overcharges have become more severe even as the overall rate of mistakes has fallen.

Now the part that argues against the easy version of this page’s claim, and which the same 2014 paper is explicit about. When its authors modelled promotional activity, only advertised sale items came out worse: they were associated with significantly greater error overall. Items on a temporary price reduction went the other way — errors and overcharges both declined significantly, which the authors state as a formal generalisation that errors are significantly less likely on temporarily reduced items. Permanently discounted items had no significant effect on error rates, and neither did items tied to store coupons.

The overcharge asymmetry on sale items does replicate outside the FTC data. The same 2014 paper obtained a separate set from Washington State — the last quarter of 2005, 100 items checked in each of 559 stores — and found the same direction there: controlling for the price of the product, relative errors were greater for sale overcharges (36.39%) than for sale undercharges (26.46%). Worth knowing how that is measured, because the wording invites a different reading: these are error sizes expressed as a percentage of the price charged, not the share of items priced wrongly.

Second, and more important: this literature examines retail and grocery checkout scanning, not restaurant point-of-sale check splitting. Nobody has run a Price Check study on split restaurant checks, and none of these findings measure what a POS does to a discount when a server moves an item. They are context for why an advertised price is a plausible weak point, not evidence about splitting. The claim that a split check strips a discount rests entirely on the Toast and Square documentation above — which is a much stronger kind of evidence for that specific claim anyway, because it is the vendors describing their own systems.

Sources: Price Check II, NIST/FTC, 1998; Price Verification FAQs, NIST; The Accuracy of Scanned Prices, Journal of Retailing 90(2), 2014; UPC Scanner Pricing Systems: Are They Accurate?, Journal of Marketing 58(2), 1994

What is actually at stake when a discount goes missing?

Percentages are abstract until they sit on a real total. Two thirds — 66% — of splitty’s US-leaning scanned restaurant receipts come to $100 or more, and nearly half — 48% — come to $150 or more (August 2026 snapshot of splitty’s own scanned receipts; percentages only, per our first-party data policy). Those are checks large enough that a percentage-off promotion is worth real money — though how much depends on the size of the discount, which these receipts do not record.

The illustration below uses invented round numbers to show the shape of the loss, not measured ones. Four people at happy hour, four discounted drinks at $6 off each, and a server who splits the check by item on a system that drops item-level discounts:

Notice what each person would have to do to catch it. Their own check is internally correct: their drink is on it, at the menu price. Spotting the loss requires remembering what the discounted price was supposed to be, or comparing against a combined check that has now been replaced by four separate ones. That is a claim about what the documents make easy to check, not a prediction about how observant any particular table is.

How to protect a discount through a split

Do these before the check closes. A closed check is a harder conversation than an open one, and on most systems it is a different operation for the staff.

What to say at the table

Short, specific and unembarrassed works better than apologetic. None of these accuse anyone of anything — they are requests about a documented system behaviour.

The structural fix: divide the receipt, not the check

The failures above are not one failure, and they do not all come from the same place — Toast removes one kind of discount, Square relocates another and moves a third along with its item. What they share is narrower: each one happens because items were moved between checks. That is the operation with a cost, and it is worth separating from the operation people usually mean when they say “split the bill.”

The catch is that headcount is exactly what people are trying to escape when one person had a $6 happy-hour beer and someone else had two $18 cocktails. The honest summary of the choice: an even payment split preserves the discount but not the fairness, and an item-level check split buys the fairness by asking the POS to rebuild the bill — which is where discounts get lost.

Splitting from the receipt is a way out of that trade. The restaurant’s check stays whole and gets paid once, so no discount is ever asked to survive being moved; the item-level division happens afterwards, on the document the restaurant already printed. The cost is real and worth naming: one card pays, so the group has to settle up between themselves rather than each paying the restaurant directly — which is the trade our comparison of splitting at the swipe versus at the receipt works through in full. What this page adds to that comparison is one specific entry: a check split at the terminal can cost you a discount you had already earned.

It is worth reading alongside how to tip when a discount is on the bill — that is the other half of the discount problem, and it runs in the opposite direction, because a discount that does survive changes what your tip should be calculated on.