Menu engineering, without the jargon.

Popularity is units sold as a share of its category. Gross profit is menu price minus the cost of the ingredients in the portion you actually serve, not the portion in the recipe you wrote two years ago. Get both from ninety days of till data. Anything shorter is a season, not a pattern. Most restaurants can pull this in an afternoon and have never done it.

Once every dish has both numbers, plot them against your category averages. Four groups fall out, and each has a different action. The mistake is treating the whole menu with one strategy.

The classification is arithmetic and arithmetic has no judgement. A slow, low-margin dish might be the only thing a regular's partner will eat, and losing it loses the table. The same goes for the one vegan option or the plain option for children. Run the numbers, then apply a sense check on which dishes are load-bearing for a group booking. Cutting the menu purely on the spreadsheet is how restaurants lose parties of six.

Long menus cost you three times over: more stock to hold, more waste, and slower decisions at the point of ordering. Cutting a menu down tends to raise the average order rather than lower it, because people choose faster and add a side instead of deliberating. If you cut, do it in one visible go rather than quietly dropping items, and tell people what stays.

Position does real work, particularly online where most ordering now happens. The first two items in a category and anything visually separated get disproportionate attention. Put a star or a puzzle there, not a workhorse that was going to sell anyway. On your own ordering site you can go further and reorder by margin rather than by tradition, which is not something a printed menu lets you do.