One taste vector. Two decisions.
The same match ranks what a guest is likely to enjoy and tells the offer engine when a discount would be wasted.
Mongolia-first · QR menu intelligence
Your QR menu is a storefront. Bestshot makes it earn.
The menu adapts inside each session, ranking dishes by taste, demand and the economics of the moment. Smart offers build the check while a live control measures what Bestshot actually added.
Every restaurant gets one honest number back: how much more the menu earned against a control it can trust.
MenuAdaptive ranking
OffersUplift measured
EvidenceBilling-grade A/B
The engineering underneath
A menu engine built to make the numbers honest.
Bestshot is a recommendation and pricing engine whose measurement system is designed alongside the guest experience.
One embedding, two brains
What they will love, and the deal worth making.
One taste vector powers the ranked menu and the offer engine, avoiding discounts on dishes the guest would already choose.
Uplift, not margin
Discount only what the offer can cause.
The engine optimizes incremental profit against a live holdout instead of spreading blanket markdowns across the menu.
Measurement is the product
A number the restaurant can be billed on.
Uplift is computed at the table-day, tested against control and logged from day one so the counterfactual remains intact.
ProPASH · in the lab
Price the rush, not the clock.
Peak and trough come from live load. At a full house the engine can rank by profit per seat-hour. At a slow hour it does nothing.
Cold-start · Mongolia-first
Useful before the first order.
Local priors such as weather, classic pairings and Tsagaan Sar give a new menu a sensible starting point.
Fairness in the schema
Personal suggestions. One table price.
Guests at one table may receive different suggestions, but never different sticker prices. The data model enforces it.
What it means for your restaurant
The upside is yours. The risk is ours.
Bestshot is priced and operated so the restaurant sees measured value before the success fee begins.
Pay from the profit we add.
A modest base fee plus a monthly-capped share of measured incremental profit. If the engine does not grow the check against control, there is nothing to share.
A number the restaurant can trust.
Each month brings a signed report: average order value against a live control, statistically tested using the same method the pricing is built on.
Live the day you sign.
No hardware or POS surgery. Print a QR per table and the menu runs in the guest's browser.
Run it in 30 seconds.
Edit menu, price and photos yourself. Mark a dish sold out once and remove it from every table's QR.
AOV without a star waiter.
Smart upsells and per-person bundles build the check on every table and every shift.
See the menu the way an analyst would.
Read margin against volume, real rush hours and the combinations guests naturally buy together.
Prove it before you pay.
Pilot restaurants run free while the measurement builds. Commit once the uplift is on the table, in writing.
Controlled simulation · live engine code
We replayed the service, table by table.
Across a controlled simulation of an izakaya's evenings, Bestshot lifted average check against a plain house menu. The result uses the same table-day, holdout-controlled method intended for billing.
Directional simulation result. Not a guarantee of production performance.The long game
One menu network. Your regulars, everywhere.
As more venues join, a guest's taste can travel with them. Each restaurant inherits demand signals no single venue could build alone.