Almost every restaurant runs a staff tasting when the menu changes. Almost none of them produce a floor that can describe the new dishes a week later. The tasting happens, everyone enjoys it, and by the following Friday the servers are reading the menu to guests again.
The failure is in the format, not in the staff. Here is what to change.
Why the standard tasting does not stick
The usual shape is: chef plates the new dishes, talks about each one for two minutes, staff eat, everyone says it is delicious, service starts. Four things go wrong.
Nobody has to retrieve anything. Listening and eating are both passive, and passive exposure produces recognition rather than recall, which is the central finding of the research on retrieval practice. Staff leave feeling they know the dishes and discover at the first table that they do not.
The information arrives in the wrong shape. Chefs describe food the way chefs think about it: technique, sourcing, inspiration. Guests ask about size, richness, what it comes with, and whether it has dairy in it.
It is a single event. One exposure, however good, does not survive a week without repetition, and the spacing effect is the reason.
And nothing is written down, so anyone who was not there, which is always somebody, gets the menu and a shrug.
| What usually happens | What guests need |
|---|---|
| Chef explains technique | What it tastes like, in plain words |
| Provenance story | What comes with it |
| Everyone tastes and nods | Someone says it back |
| Nothing written down | Cards or a sheet for whoever missed it |
| One event | Repetition across the week |
The format that works
The change is small and it costs no extra time.
Taste first, talk second. Let people eat a spoonful before anyone describes it. Tasting while being talked at splits attention and neither half lands.
Then have the floor describe the dish, not the chef. Go around and ask two or three people to give their one-sentence description. This is the entire mechanism: it converts a passive session into retrieval practice, and it surfaces the misunderstandings immediately rather than at a table.
Then the chef corrects and adds the two or three facts the floor missed, which are almost always size, accompaniment and an allergen.
Then write it down, in the floor’s words rather than the menu’s. The sentence a server produced under mild pressure is the sentence that will come out at a table.
Twelve dishes takes about thirty minutes in this format, which is roughly what the passive version takes anyway.
What to capture for each dish
Six fields, and they are not the menu copy.
The one-sentence description, in plain spoken language, built the way a one-sentence dish description sets out.
What it comes with, or explicitly that it comes alone. This prevents more complaints than anything else on the list.
Size relative to the rest of the menu.
Flavour direction in one or two words: rich, sharp, mild, smoky.
Allergens and whether modification is possible, confirmed rather than assumed. UK food businesses must provide allergen information, and the tasting is the cheapest moment to get it into everyone’s head at once, alongside the allergens every server should know.
Preparation time, if it is unusual.
That set fits on a card, which is the practical reason to capture it that way rather than as a paragraph. MenuFlashcards can build the deck from a photo of the new menu in a couple of minutes, and the tasting is where the back of each card gets filled in with the answers the floor just worked out together.
Running it when you cannot feed everyone
Full tastings are expensive and many kitchens cannot plate twelve dishes for fifteen staff. The format survives at smaller scale.
Taste the components rather than the dishes. A spoon of each sauce, a piece of each protein, a bite of each garnish. This is a fraction of the cost and it covers most of what the floor needs to describe.
Split it across days. Three dishes at a time, before service, for four days. This is better than one big session anyway, because it spreads the repetition out.
Taste the difficult ones only. The dishes people will ask about, the ones with unusual textures, the ones that sound like something they are not. Half the menu is self-explanatory.
Use the plates that come back. A dish returned for a non-food reason is a free tasting, and most kitchens will let the floor try it rather than binning it.
| Constraint | Workable version |
|---|---|
| No budget for full plates | Taste components: sauces, proteins, garnishes |
| No time before service | Three dishes a day across a week |
| Split shifts, nobody together | Run it twice, same script |
| Large menu | Taste only the dishes guests ask about |
The follow-up that makes it stick
The tasting is the input. Without a second and third exposure, most of it is gone within days.
Ask one dish at the next three pre-shifts. Spreading the repetition out is what makes it stick, since distributed practice beats a single session by a wide margin. Not a lecture, one question to one person, thirty seconds. “Describe the bream.” This is the cheapest retention mechanism in the building and almost nobody uses it.
Put the six fields where people can drill them. A deck on a phone gets used on a break; a laminated sheet in the office does not.
Ask new starters to describe the specials before service rather than telling them. Same mechanism, applied to the people who need it most.
And revisit after a week. The dishes that come back wrong at day seven are the ones to re-taste, and there are usually only two.
Common mistakes managers make
Running the tasting during service prep, when nobody can attend properly.
Letting the chef talk for twenty minutes, which is enjoyable and produces almost no floor knowledge.
Skipping the allergen detail because it feels like a compliance topic rather than a tasting topic. It is the single most important field on the card and the tasting is the best moment to fix it in people’s heads.
Not inviting the bar, or not inviting the kitchen porters who run food. Anyone who talks to a guest about a dish should be there.
And treating it as a treat rather than as training, which quietly tells everyone that remembering it afterwards is optional.
What to measure
You do not need a quiz. Ask three servers on the Friday after a Monday tasting to describe two dishes each, and listen for the six fields.
If the descriptions are accurate and include what the dish comes with, the tasting worked. If they are reciting menu copy, it did not, and the fix is the retrieval step rather than a longer session next time.
Tasting for the bar, which usually gets skipped
Drinks lists change as often as food menus and almost never get the same treatment, which is why bar knowledge on most floors is thinner than food knowledge.
Run the same format. Taste, describe, correct, capture. Cocktails are easier than food here because the components are consistent and the descriptions are shorter.
Capture five fields per drink: what it is built on, whether it is sweet or dry, whether it is served long or short, the garnish, and whether there is a non-alcoholic version. That last field is asked about constantly and answered badly almost everywhere.
Include the wine list in some form, even if a full tasting is impossible. Two or three honest food pairings per section, agreed by the floor and written down, is worth more than a sommelier talk nobody retains.
Tasting with a new menu versus a rotating one
The format is the same and the cadence is not.
A full menu change is a project. Run the tasting a few days before launch rather than on the morning, so there is time for the follow-up questions and for the two dishes that come back wrong.
A rotating specials board needs a smaller daily version, which is really learning daily specials as a habit: one dish, one taste, one description said back, thirty seconds at the pre-shift. Done every day, this is more effective than any quarterly event.
Seasonal swaps sit in between. Treat the three or four changed dishes as a mini tasting rather than assuming the floor will absorb them from the printed menu.
| Menu change | Tasting cadence |
|---|---|
| Full menu launch | Full session, a few days before launch |
| Seasonal swap of a few dishes | Mini tasting of the changed dishes only |
| Daily specials | Thirty seconds at each pre-shift |
| A single new dish | One taste, one description, at handover |
The cost argument, for whoever holds the budget
Tastings get cut because they look like a food cost with no return. The return is real and it shows up in three places.
Fewer sent-back dishes, because the floor described the dish accurately and the guest got what they expected. Every returned plate costs more than the tasting portion did.
Higher attachment, because servers who know what a dish comes with sell the side without pushing, and servers who know what is in a cocktail sell the second one.
Faster new starters, because a tasting plus a written card set is the difference between a server who is useful in week one and one who is useful in week four.
None of that requires an elaborate event. The version described here costs a few components and half an hour, and the follow-up costs thirty seconds a day.
Who should run it
Not always the chef, which surprises people.
The chef has to be there for the correction step, because they are the only reliable source for allergens and composition. They are frequently the wrong person to run the session, because the format asks the floor to talk more than the kitchen does, and that is easier to hold when a manager is chairing.
Give someone the job of writing it down, and make it someone who will actually be on the floor. Notes taken by a person who then serves the dishes are noticeably better than notes taken by whoever was free.
And have one person own the follow-up questions at the next three pre-shifts, so that step does not quietly disappear the way it usually does.
Frequently asked questions
How do you run a staff menu tasting that people remember?
Have staff taste first, then describe the dish themselves in one sentence before the chef speaks. That single change converts a passive session into retrieval practice, which is what produces recall rather than a vague sense of familiarity. Then capture the floor’s own words in writing and ask about one dish at each of the next three pre-shifts.
What information should a menu tasting capture?
Six fields per dish: a plain one-sentence description, what it comes with or that it comes alone, size relative to other dishes, flavour direction in a word or two, confirmed allergen information and whether it can be modified, and preparation time if unusual. That is a different set from the menu copy and from what chefs naturally describe.
How do we run a tasting when we cannot afford to plate everything?
Taste components rather than finished dishes: a spoon of each sauce, a piece of each protein, a bite of each garnish. That costs a fraction as much and covers most of what the floor needs to describe. You can also split the tasting into three dishes a day across a week, which spreads the repetition and works better anyway.
Why do staff forget the menu a week after the tasting?
Because a tasting is a single passive exposure, and passive exposure produces recognition rather than recall. Staff leave feeling they know the dishes and discover at the first table that they cannot produce the description. Repetition across the following days, in the form of one question at each pre-shift, is what fixes it.
Who should attend a menu tasting?
Anyone who talks to guests about food, which usually includes the bar, food runners and hosts as well as servers, and is wider than most restaurants invite. If shifts make one session impossible, run the same short script twice rather than letting half the floor learn the menu from the printed card.
