Menu engineering — the practice of analysing each dish on a menu against its profitability and popularity to determine whether it deserves its place — is not a new idea. It has been a standard element of hospitality management education since the 1980s. What is changing is the quality and accessibility of the tools available to do it.
A handful of UK restaurant operators, ranging from independent single-site businesses to mid-scale groups, have begun using AI-powered menu analytics platforms that go significantly beyond the traditional four-box matrix of stars, ploughhorses, puzzles and dogs. The results they are reporting challenge several assumptions about which dishes on a menu are actually earning their keep.
What the Tools Do
The new generation of menu engineering software integrates with a restaurant's POS system to pull granular sales data — not just how many portions of each dish were sold, but when, at what table size, alongside which other dishes, in which combination with beverages, and in which weather conditions. Against this, the platform layers food cost data entered by the operator to calculate actual contribution margin per dish per service.
The AI element comes in pattern recognition: identifying correlations and interactions that a human analyst looking at a spreadsheet would not catch. Which dishes are being ordered by tables that then order the most wine? Which desserts increase check average without cannibalising the table time needed to turn covers efficiently? Which popular starters are being ordered by guests who then spend least on mains?
"We thought our Sunday roast was our best commercial product," said the owner of a 60-cover gastropub in the South East who trialled one such platform earlier this year. "The data showed it was our most popular and our worst margin item, and that tables that ordered it were drinking significantly less than tables on the regular menu. We'd been subsidising Sunday trading for two years and genuinely didn't know."
The Waste Connection
The food waste dimension of menu analytics is proving particularly valuable for operators who have committed to sustainability targets but have not previously had the data infrastructure to track waste accurately at dish level.
By correlating prep volumes — entered by kitchen management — against sales data, the platforms identify where over-preparation is occurring systematically. The findings typically reveal a small number of dishes responsible for a disproportionate share of waste: usually specials that were ordered less frequently than anticipated, or garnish-heavy dishes where the garnish is prepared in advance and frequently discarded unused.
One London group operating four sites reports reducing food waste by 23% within six months of implementing an AI menu tool, driven primarily by changes to prep volumes for the eight dishes that the system identified as the group's worst waste offenders. The financial impact — calculated against the cost of the wasted ingredients and the reduced skip collection volume — was approximately £18,000 across the estate over the period.
The Menu Restructuring Question
The most commercially significant output from AI menu engineering is typically a recommendation to remove or significantly redesign a number of dishes. This is where the technology meets operator psychology — and where the gap between what the data suggests and what operators are willing to do is widest.
Every restaurant has dishes on the menu that are there because the chef is attached to them, because a regular guest orders them every week, or because they represent an identity statement that the operator does not want to abandon regardless of its P&L contribution. The AI tool has no view on any of this. It sees contribution margin and frequency, and it makes its recommendation accordingly.
The operators who are getting the most from these platforms are those who treat the data as one input among several rather than a directive. The menu engineering analysis tells you which dishes are working commercially. It does not tell you which dishes are worth keeping for reasons the data cannot capture. The judgement about where those two things intersect is still a human decision.
"The tool told me to remove our wood-roasted cauliflower," said one chef-owner. "It had a low margin and moderate frequency. What it couldn't tell me was that it is the dish every food writer who visits mentions, that it is why our Instagram exists, and that it drives bookings worth far more than the dish's contribution margin. We kept it. But we repriced it."
Platforms in This Space
Several platforms are now positioning specifically for UK restaurant operators in this category, including UK-developed tools from hospitality software companies and international platforms including MarketMan and Apicbase that have established UK customer bases. Integration quality with common UK POS systems — Lightspeed, Epos Now, Square — varies and should be assessed carefully before committing to any platform.
The cost of entry has fallen significantly over the past two years. Platforms that were pricing at enterprise level in 2022 are now offering entry-level tiers accessible to independent operators at under £150 per month — a price point at which the ROI from even modest improvements in menu margin or waste reduction is straightforward to justify.
For any restaurant carrying more than 25 dishes on its main menu, the probability that AI menu analytics will find at least one material insight that a manual review would miss is, based on the operator accounts we have gathered, close to certain.