Sparrows/Bestshot
AI menu ranking and offers for restaurants, bars and lounges

The menu that knows the table.

The QR code on your table is the only staff member who talks to every single guest. Right now it shows a PDF. Bestshot turns it into a menu that ranks itself to the person holding the phone, offers the right add-on at the right moment, and reports back at the end of the month with a number you can check.

15 dishes ranked per guestFull price on the anchor

The table

Scan, rank, offer.

Three moments. The guest sees a better menu and you see a check you can measure.

The QR code becomes the menu's best waiter.

It ranks, suggests and measures without hiding the full menu or changing a price from one guest to another.

01

Fifteen dishes, ranked for this person

The menu reorders itself to their taste, today's weather, the time of day and what the table has already ordered. “All dishes” stays one tap away, always. Curated, never hidden.

02

Or let the menu choose

Tired of reading sixty dishes? One button. It is not random: it picks inside their taste, inside their price range, and skips anything the kitchen has run out of.

03

The offer fires on the anchor

When something goes in the cart, Bestshot looks for an add-on worth offering. The dish they already want stays full price. Only the item they would not have ordered gets a discount, and only when the maths says it pays.

The rules

Four rules we do not break.

Discounting is easy to do badly. These are the constraints, and they are in the code, not the brochure.

01

Never discount a sure thing

If the system can already tell someone wants a dish, it will not hand them money off it. The offer engine only spends on purchases it can actually cause.

02

Personalize what, not how much

The bundle can differ per guest. The discount percentage cannot. Two people at one table never see two different prices for the same dish.

03

Never below cost

Depth is capped against the item's margin at a single chokepoint, enforced in the database, not by whoever is configuring the promo that day.

04

Never drive volume you cannot serve

At peak, discount depth drops to near zero. In a quiet hour it opens up. The system pushes demand into the gaps in your evening, not on top of the rush.

The proof

Half your tables run without us. That is the point.

Every table, every day, is assigned to Bestshot or to the normal menu. At the end of the month you get one number: what the Bestshot tables spent, against what the others did. If the number is not there, we would rather you saw that too.

MENU 15 dishes ranked per guest
Assigned by table-day Not by session

Four people sharing one bill would otherwise contaminate both sides of the test.

Cluster-robust Analysed at the level it was assigned

Session-level significance on clustered assignment produces confident nonsense.

Every impression logged What was shown and what happened next

Orders alone cannot tell you what someone saw and skipped, and that data cannot be recovered later.

Bestshot

We are in pilot, and we publish the number either way.

Five restaurants, two months free, full access, and a control group running the whole time. In exchange we use the data and publish what happens. If you run a room with enough tables to measure, we would like to talk.

Runs on Vida, the engine shared by all four Sparrows products.

MenuRanked per guest
OffersOnly when they pay for themselves
PricingWeighted to measured lift
ProofA control group, always running