Restaurants spent decades guessing which dishes to promote. Menu engineering made it a discipline; digital menus made it testable. Here’s the classic playbook, updated for screens.
Know your stars from your plowhorses
Classic menu engineering sorts dishes by popularity and margin: stars (popular, profitable — promote them), plowhorses (popular, thin margin — reprice or re-portion), puzzles (profitable, ignored — rename, photograph, reposition), and dogs (neither — let them go). On paper this took quarterly analysis; a digital menu shows you the data continuously.
Position is persuasion
Eyes land first on a list’s opening items and remember the last — bury the plowhorse mid-list, open sections with stars. On a phone screen, “first” means before the scroll: your best margins belong in the opening viewport.
The small print that moves money
Dropping currency symbols (“245” not “245 Kč”) measurably softens price pain. Decoy anchoring works too: the premium steak nobody orders makes the second-priciest feel reasonable — its job is to be seen, not sold.
Photos: selective, not total
Photographing everything flattens the menu; photographing your stars and puzzles points appetite where margin lives. Pair it with descriptions that sell provenance — “farm” and “grilled over oak” outsell adjectives like “delicious” every time.
Digital menus like ProMenu.cz make every one of these testable weekly instead of yearly. Ask us to engineer your menu — the data is already in your orders.
Frequently Asked Questions
Sorting dishes by popularity and margin — stars, plowhorses, puzzles, dogs — and designing the menu to promote the profitable ones.
Yes — prices without currency symbols measurably soften price pain; it is one of the oldest verified menu tricks.
Where eyes land first — section openings, and on phones the first viewport before any scrolling.
No — photograph stars and puzzles; photographing everything flattens attention instead of directing it.
A premium item priced to make the next tier feel reasonable — its job is to be seen, not ordered.
The data is continuous — you can test names, photos and positions weekly instead of guessing quarterly.

