Average Order Value: What It Is and How to Actually Move It

Average order value is exactly what it sounds like: total revenue divided by the number of orders, giving you a single number that represents how much a typical table spends. It's one of the easiest restaurant metrics to calculate and, for most owners, one of the most frustrating to actually influence on purpose, since the obvious levers, telling staff to upsell more, tend to move it less than expected.

Why the obvious moves often disappoint

Asking servers to suggest more add-ons sounds like a direct path to a higher ticket, and sometimes it works, but often it just adds friction to the ordering experience without meaningfully changing what guests actually buy. A generic upsell pitch, delivered without regard to what a specific table has already ordered, reads as exactly what it is: a sales tactic rather than a genuine recommendation. Guests are reasonably good at detecting the difference, and a pitch that feels like a script tends to get politely declined regardless of how good the suggested item actually is.

Blanket price increases have a similar problem. They raise average order value on paper immediately, but if they're not grounded in what specific dishes can actually bear a higher price without hurting satisfaction, they risk quietly depressing order volume in a way that erases the gain, or worse, damages loyalty among regulars who notice.

What actually tends to move the number

• Relevant, specific pairings suggested at the right moment, not a generic add-on pitch to every table
• Well-built combos that make ordering more items feel like a single easy decision rather than several small ones
• Targeted price adjustments on dishes with proven room, not across-the-board increases
• Personalizing what's suggested based on what a specific guest has already ordered or tends to like

Why relevance beats volume of suggestions

The through-line across everything that actually works is relevance. A suggestion that genuinely fits what a guest already ordered gets accepted far more often than a generic one, and it does so without making the guest feel sold to, which matters just as much for repeat business as it does for the ticket in front of you right now. The goal isn't suggesting more things. It's suggesting the right thing at the right moment, which requires actually knowing something about the guest and the dish rather than running the same pitch on every table regardless of context.

This is where personalization consistently outperforms a blanket sales push, because it replaces a guess about what might sell with an informed suggestion based on real signal about what this particular guest is likely to want.

Why the same pairing doesn't work for every table

A pairing that works beautifully for one table can fall flat for the next, even if they ordered the exact same main dish. A table that already ordered two appetizers and a drink each is a poor target for another add-on suggestion, no matter how well it pairs on paper, since they're already at a comfortable order size. A table that ordered lightly might genuinely welcome a relevant suggestion. Generic upsell rules that ignore this context end up pitching the same add-on to both tables, which is part of why blanket upsell scripts underperform something that actually accounts for what's already on the ticket.

Measuring whether a change actually worked

Average order value can be moved briefly by a one-off promotion or a pushy server shift without reflecting any real, lasting change in guest behavior. The more reliable way to check whether an approach is working is to track it over several weeks rather than a single busy night, and to watch whether it holds up across different days and different server shifts, not just the one night someone happened to try harder.

TasteGraph's approved illustrative data shows an example 11.4% lift in average order value driven by exactly this kind of relevant, per-guest pairing surfaced through the menu itself, rather than a server working from a generic script every table hears the same way.