ShiftTrained Data Report · No. 1
The Price Blind Spot
The median restaurant menu has 50 dishes on it and about half of them contain an allergen. We measured what staff actually know about all of it. The worst answer on the menu turned out to be the price.
Terry Psaltakis, Founder, ShiftTrained · Published August 16, 2026· Live menu and training data, 10 April to 16 August 2026
I spent thirty years on restaurant floors before I built any of this, and in all that time I never once saw anybody actually measure what their staff knew. We quizzed them, sure. Somebody holds up a plate at pre shift and asks what is in it, three people answer, service starts, everybody forgets about it.
So we went and counted. First what is on the menus, because you cannot say much about what a server knows until you know how much there is to learn. Then what they got right and wrong.
Part One
What is actually on the menu
Every dish we have parsed across 100 restaurants. 11,318 items.
50
dishes on the median restaurant menu
The average is 75, but one menu in the set runs to 597 items and drags the average around, so 50 is the honest number. That is what you hand a new server in their first week. Usually a printed copy and two shifts following somebody around.
49.4%
of dishes carry at least one allergen
On a 50 item menu that works out to about 25 dishes a server has to get right the moment a guest says the word allergy. There is no version of that answer where close enough is good enough.
Share of all 11,318 items. Gluten n=3,166, dairy n=3,052, eggs n=1,308, sulfites n=685, fish n=537, shellfish n=445, nuts n=403.
Gluten and dairy each show up on about one dish in four. That is not something a server looks up when it comes up. That is most of your menu.
8.3%
of dishes are marked vegetarian, and 3.6% vegan
Gluten free sits at 3.2%. Read the trade press and you would think every kitchen in the country has gone plant forward by now. The menus say something different. More than eight dishes in ten carry no dietary label at all. Guests are asking for this. The menus have not caught up yet.
Part Two
What staff actually know
38,079 answered training questions from 206 restaurant staff.
37.9%
of price questions were answered wrong
Ingredient and prep questions came back at 19.4%. Every restaurant trains on the food. You walk the new server through the dishes, what is in the sauce, what comes on the side. Nobody drills prices, because everybody assumes prices are the easy part.
They are not. Your staff know what is in the dish about twice as well as they know what it costs.
Miss rate by topic. Questions sorted by their text. Price n=639, wine n=3,859, bar n=1,854, allergens n=398, ingredients n=4,479.
4 of 4
restaurants showed the same price gap
One restaurant with a confusing menu could produce that headline by itself, so we split it out. Every restaurant with enough price questions to measure came back the same way, between 1.8x and 3.0x its own ingredient miss rate.
| Restaurant | Price miss | n | Ingredient miss | n | Ratio |
|---|---|---|---|---|---|
| A | 47.1% | 87 | 24.3% | 341 | 1.9x |
| B | 44.6% | 157 | 14.8% | 969 | 3.0x |
| C | 36.9% | 141 | 20.3% | 1,244 | 1.8x |
| D | 34.1% | 217 | 15.3% | 583 | 2.2x |
The thing your staff know worst is the thing that decides your check average.
30.1%
of wine questions were answered wrong
The second blind spot, and on much heavier volume at 3,859 answers. Producer and region questions were the weakest of the lot. If you are selling bottles, that is margin walking out the door every time a server gets stuck and points at the house pour instead.
84.4%
Final score of the staff who did worst on day one
Staff who scored badly enough to retake started at 67.2% and finished at 84.4%. That is an average gain of 17.2 points, and 241 of 293 improved.
The obvious objection is that retakers are just the strugglers. They are. Staff who passed first time and never came back averaged 78.5%. But that is exactly what makes this worth printing. The people who started furthest behind did not just catch up. They finished 5.9 points ahead of the ones who never needed a second go.
| Group | First score | Final score | n |
|---|---|---|---|
| Retook at least once | 67.2% | 84.4% | 293 |
| Passed first time | 78.5% | 78.5% | 189 |
Counted per person per quiz, so somebody who sat two different quizzes appears twice.
88
seconds is the median time a server spends on a quiz
That is answering time, not elapsed time. The median question takes 7 seconds and only 1.09% ever run out the 20 second clock. Staff are not refusing to sit still for this. They answer in seconds, standing up, between tables. Menu training does not fail because your people will not do it. It fails because nobody asks them.
89%
of staff eventually reached their restaurant's pass mark
430 out of 482 got there. Just over half, 249, passed on the first attempt. Another 92 passed on the second and 41 on the third.
52 never passed at all. That number belongs in here as much as the good one does. This is not a machine that hands everybody a certificate at the end, and anybody selling you one that does is selling you paper.
What this is, and what it is not
Limitations
- These are our customers, not a random sample of restaurants. A place that goes out and buys menu training is probably not the same as a place that does not, and the menus skew toward independents and small groups rather than chains.
- Allergen tags are pulled out of the menu text by AI, not verified in anybody's kitchen. They reflect what the menu says or clearly implies. Where a menu simply does not mention it, we undercount.
- The training half rests on the 8 restaurants whose staff have sat quizzes so far, a much smaller group than the 100 whose menus we parsed. The menu numbers are the broader of the two and should be read that way.
- Questions were sorted by their text, not by a taxonomy we built by hand. A question naming both a price and an ingredient counts as price.
- The allergen question figure rests on 398 answers. That is the thinnest number in here and I would not lean on it without more.
- Price figures are dollar denominated only. The corpus includes restaurants pricing in Rand, Naira, Rupees and Ringgit, and mixing them would make any average meaningless.
- Every number is reproducible from the underlying records. We will publish the same measures every quarter whether or not they keep making us look good.
Cite this
Psaltakis, T. (2026). The Price Blind Spot: what is on restaurant menus, and what staff actually know about it. ShiftTrained Data Report No. 1. 11,318 menu items across 100 restaurants and 38,079 training answers from 206 staff, 10 April to 16 August 2026.
Published under CC BY 4.0 — quote it freely with attribution. The living versions of these figures are on menu statistics and training statistics, which refresh daily.
About the Author
Terry Psaltakis is a 30-year restaurant operator who has opened more than 20 concepts across multiple markets, in every role from dishwasher to Owner. He founded ShiftTrained in Chicago to solve a problem he lived for three decades: pre-shift meetings don't actually train staff. Terry writes about the operational side of restaurant training, AI in hospitality, and what works on the floor.
LinkedIn · terry@shifttrained.com
Last reviewed June 2026
“Since we started using ShiftTrained, wine sales for both bottle and by-the-glass are up 34%. The staff is not scared to talk about the wine anymore.”
George G. · Black Barrel · Chicago
Ready to train your staff BEFORE service?
Upload your menu. AI builds the quiz. Your team takes it on their phones. Free trial, no credit card.
Free trial · No credit card