Does tabletop ordering increase check size? What research shows

Barely. The strongest study, peer reviewed and based on 2.6 million transactions in a 66-restaurant US chain, found tabletop ordering raised sales per check by “approximately 1%” and cut meal duration “by close to 10%”. The popular claim of 20% to 30% bigger checks matches a hypothetical example in a 2016 report and vendor marketing, not a measured result. The real gain is speed.

Where does the “20-30% bigger checks” claim come from?

We could not find an independent study that supports it. We did find two likely sources: a worked illustration that later got repeated as a finding, and vendor figures published without method.

The hypothetical. A 2016 Cornell Hospitality Report on tabletop devices set the scene with this line: “if the tabletop technology can improve the average guest check by just $3, that would result in a 20-percent improvement considering a restaurant average check of $15” (Susskind and Curry, Cornell, 2016, p. 4). It is an “if”, written to show why the question matters. It is not a result.

Vendor figures. These are real quotes from vendors’ own pages. None gives a sample, a time period or a comparison group:

• Square (2021): “Businesses using Square that have enabled open tabs on Self-Serve Ordering see 42% higher average ticket sizes” (Square, 2021). Note that this covers open tabs only.
• Butlaroo, which sells an AI waiter, says as of October 2026: “30% lift in average cover value vs. traditional service” (Butlaroo).
• AI-Waiter.com says as of October 2026: “UP TO +25% AVG CHECK INCREASE” (AI-Waiter.com).

A vendor figure is not necessarily wrong. But without a method you can’t tell whether it compares like with like, or simply shows that bigger spenders like the feature.

What did the peer-reviewed study actually find?

About 1% more per check, about 10% shorter meals, and about 11% more sales per minute. The meal-time effect is ten times the check-size effect.

The study is Tan and Netessine, “At Your Service on the Table”, published in Management Science in October 2020 (abstract). A US casual-dining chain added tabletop devices to its restaurants at different times, which let the authors compare restaurants before and after, against those not yet switched over (a “difference-in-difference” design). The published abstract says:

“We find that the tabletop technology is likely to improve average sales per check by approximately 1% (95% confidence interval is from 0.8% to 1.02%), and reduce the meal duration by close to 10% (95% confidence interval ranges from −9.94% to −9.54%). The combination of these two effects increases the sales per minute or sales productivity by approximately 11%.”

It also says “the technology does not change the staffing level.”

Three details worth knowing:

1. The earlier version said 2.91%. The 2017 working paper estimated 2.91%, “approximately $0.88 of the average check size of $30.45”. The peer-reviewed version settled on about 1%. If you see “3%” quoted, it is the draft.
2. Servers still took the main order. In that chain, “the restaurant chain requires waiters to take customers’ food and first alcoholic beverage orders”. Guests used the device mostly to reorder drinks and to pay.
3. Drinks moved most. The working paper found alcoholic drink items rose 13.83% per check, while food items rose 1.76%. Easy reordering, not bigger meals, drove most of the small lift.

What about the Cornell study’s 9% higher checks?

It is real but weaker evidence. In 45 restaurants of one chain in January 2016, guests who ordered an entrée on the tabletop device spent $43.76 against $40.15 for those who didn’t, $3.61 more (Susskind and Curry, 2016, p. 8).

The catch: guests chose whether to use the device. People who like gadgets, order more courses or stay for drinks may be exactly the ones who pick up the tablet. The report itself says device use “is associated with higher check averages”, which is the right wording. One co-author was a data scientist at E la Carte, the company that made the device.

The same report’s time findings point the same way as the peer-reviewed study:

• Guests who paid on the device had a table-turn time of 45 minutes 21 seconds, against 53 minutes 45 seconds for those who didn’t, a “15.5-percent reduction” (p. 7).
• In one restaurant, server time per table fell 36% when guests paid by device and 65% when they ordered and paid by device. That was from 63, 33 and 7 tables, so treat it as a hint (p. 6).

Side by side: what each source shows

Does this apply to QR ordering on guests’ phones?

Not directly. Both studies looked at tablets fixed to the table in US casual-dining chains, ten years ago or more. We did not find a peer-reviewed study of check size for ordering from guests’ own phones, or for chat-based AI waiters. Phones and chat may do better or worse; nobody has published the answer we could verify.

What carries over is the mechanism: when guests don’t have to wait for a server to order another round or to get the check, meals get shorter.

What does 1% vs 10% mean for one restaurant?

Worked example (fictional, with stated assumptions). Casa Olivo has 20 tables. On Friday its three peak hours (180 minutes) have a waitlist the whole time. A table stays 90 minutes on average and spends $80.

Suppose, only for this example, that the study’s effects applied to Casa Olivo:

Two lessons. The time effect is worth about ten times the check effect. And it is worth something only when guests are waiting for tables: on a quiet Tuesday with empty tables, shorter meals add nothing. These are illustrative numbers, not a forecast and not a WaiterPilot result.

How do you measure it in your own restaurant?

Measure before you switch, and compare like with like. Here is a worksheet.

1. Pick the weeks. Four normal weeks before, four after. Skip holidays and local events.
2. Track per service, from your POS and your floor notes:

3. Compare Friday with Friday, not Friday with Monday.
4. Don’t compare users with non-users and call the difference an effect. That is the Cornell catch: the guests who choose the new tool are different.
5. Write down what success means before you start, for example “two more tables seated on Friday peak”.

What does WaiterPilot claim?

Nothing on check size. WaiterPilot asks guests about choices such as sides and cooking, and after a guest adds a dish it may ask once whether they’d like anything with it. It confirms the order before it goes to staff, but we have no published results on check size and we won’t quote a number we can’t support.

What we do say: guests order by chat in their own words and language, the order reaches the staff board without anyone walking to the table, and the price is a flat $200 per location per month after 30 days free, with 0% commission. There is no POS link, so your POS stays the source for the sales figures in the worksheet above. Read how QR table ordering works with WaiterPilot, or how the cost compares in flat fee vs commission.

Your next step

Before you pick any ordering tool, fill in the “before” column of the worksheet for four weeks. Then try a sample order and judge WaiterPilot on what it can change: how fast an order reaches your staff, and how much time servers get back.