A Beginner's Guide to Menu Personalisation for Restaurant Owners

If you're hearing the term "menu personalisation" for the first time and it sounds vaguely complicated or expensive, the good news is the underlying idea is simpler than the phrase makes it sound. This is meant as a plain starting point, not a sales pitch.

The basic idea in one paragraph

Your menu, the dishes, categories, and prices, stays exactly the same for every guest. What changes is the order dishes appear in, based on what a specific guest has told you or shown you they like. A guest who's flagged themselves as vegetarian sees vegetarian dishes surfaced higher. A guest who orders spicy food regularly sees the spicier options first. Nobody sees a different menu, they see the same menu arranged around them.

How it actually learns preferences

• Guests can directly state preferences, like spice tolerance or dietary restrictions
• Past orders, if a guest is a returning customer, inform what surfaces first next time
• Quick feedback, like a thumbs up on a specific dish, refines what gets shown going forward
• None of this requires a guest to fill out a long form before they can order

What it costs guests, in effort and privacy

A reasonable concern for any owner considering this is whether it asks too much of guests or feels invasive. Done well, it should ask for almost nothing upfront, a guest can order without providing any preferences at all, and personalisation simply activates more as they interact with the menu. Nothing about it should require guests to hand over information that feels excessive for the purpose of ordering dinner.

A simple example to make the idea concrete

Say a guest visits three times over a month. On the first visit, they see the standard menu order and happen to try a paneer dish, then leave a quick thumbs up on it. On the second visit, that dish and similar vegetarian options are surfaced a little higher, not because the menu changed but because the system has one data point suggesting what this guest liked. By the third visit, if they've also mentioned a mild spice preference, the whole upper section of what they see leans toward mild, vegetarian-friendly options, without them ever filling out a form or being asked a single direct question. That's the entire mechanism, repeated at a larger scale across every guest who interacts with the menu at all.

Common misconceptions worth clearing up early

• It doesn't hide any dish from anyone, everything on the menu remains reachable, ordering is never restricted
• It doesn't require guests to create an account or log in to get any benefit
• It doesn't replace your kitchen's actual recipes or pricing decisions in any way
• It doesn't need a large volume of guests to start working, even a single returning guest sees a difference
• It isn't a one-time setup that then runs itself forever untouched, it improves with a bit of ongoing attention

What owners should expect in the first weeks

Personalisation gets better with more data, so don't expect dramatic results on day one with your very first guests. What you should notice fairly quickly is the dashboard side of things, real information about which dishes guests actually respond to, which tends to be useful well before the personalisation itself has fully calibrated. Give it a few weeks of real guest traffic before judging whether the reordering itself feels like it's working, that's usually enough time for patterns to start forming for your regulars.

It's also worth remembering that a first-time guest gets less personalisation benefit than a fifth-time regular simply because there's less history to work from. That's expected and not a flaw, the system is meant to get more useful the more a specific guest interacts with your menu, not deliver the same experience to everyone regardless of history.

TasteGraph handles this exact process: the same menu, same categories, same dishes, reordered per guest based on what they've told you or shown you, with an example lift in average order value of around 11.4% in cases where it's been measured, treated as an illustration of what's possible rather than a guarantee.