Let the data classify your menu
POS data shows what sells and, combined with food cost, what is profitable. Classic menu engineering sorts items into stars (popular and profitable), plowhorses (popular, low margin), puzzles (profitable, unpopular), and dogs (neither). Pull item sales for the last seven days and sit with the chef before you order the week, not after. A rough cost next to a real count beats guessing from whatever the pass felt busy with. Use the names servers already sell, so the menu matches the button. If a patio week is in the mix, split those weeks out before you call a cold plate a star and a colder-month dish a dog. Canadian restaurant owners use menu engineering POS data for that split, so a patio week does not write the fall menu.
Act on each category
Feature stars prominently, re-engineer plowhorses to improve margin, promote puzzles harder or reposition them, and cut or rework dogs. Small menu changes guided by data can meaningfully lift overall margin. Pick one plowhorse and one change for the week: portion, recipe, or price. For a puzzle, decide whether the name, the description, or the placement is why almost nobody orders it. Take dogs off or rewrite them, unless the count is low only because you mixed a patio week with the dining room. Check the following Monday to see if the item moved. A label you never act on leaves the menu as the printer left it last season.
Price and place with intent
Use the data to set prices that reflect value and to place high-margin items where eyes land first. The first items a guest sees should be stars for the service you are in now, not a special that sold only while the patio was open. A lunch sandwich that wins at noon can be a puzzle at dinner. Burying it on the dinner menu will not fix the lunch board. When you reprint, put the true stars where the eye lands, and do not describe a plowhorse as the best plate in the house. A price copied from a summer week will be wrong in the fall.
Find the reports you need
Caviar by Armour Payments surfaces item-level sales and trends so menu engineering is a weekly habit, not an annual scramble. See restaurant POS or book a demo. On Monday, open item sales for the last seven days and sort them with the chef before prep. You need counts, and a food cost you trust, under the names already on the screen. If the report cannot separate a patio week from a dining-room week, or lunch from dinner, the next menu will be built on the wrong service. A dashboard with no item list will not survive the first Monday you try to use it.
A Monday half-hour with last week’s sales
Monday before prep, pull item sales for the last seven days and sit with the chef for half an hour. Mark stars, the dishes that are popular and profitable, and keep them near the top of the menu. Mark plowhorses, the dishes that sell and cost too much, and pick one change for the week: portion, recipe, or price. Mark puzzles, good margin and almost nobody orders them, and decide whether the name, the description, or the placement is the problem. Dogs come off or get rewritten. Use the same list the servers already sell from, so the words on the menu match the button on the screen. Caviar’s item-level sales make this a weekly habit. A rough cost next to a real count beats guessing from whatever the pass felt busy with. Those reports live in the restaurant POS.
Patio items that only look like stars
A summer special can dominate the report for weeks and then teach you the wrong menu for fall. If you rank dishes on July volume alone, the cold plate and the patio drink look like stars, and the dish that actually carries a Tuesday in November looks like a dog. Split the report by the weeks the patio was open versus the weeks it was not, before you delete anything. Do the same for daypart. A lunch sandwich that wins at noon can be a puzzle at dinner, and burying it on the dinner menu will not fix the lunch board. Price and placement should follow the service you are in right now. When you reprint, put the true stars where the eye lands first. Book a demo to see item trends by week.