Values that forbid something.
These aren't a mission-statement exercise. Each names something the company actually does day to day, paired with the checkable mechanism behind it — because a value that doesn't forbid anything isn't a value. It's a slogan.
A values page is usually the easiest place on a company's site to write something unverifiable — a stated value sounds true by construction. So this page holds itself to a harder test: every practice below links to the specific page where its mechanism is proven, in enough detail that you can go check it directly instead of trusting a sentence that merely sounds right.
Each practice, engraved with its proof.
We describe what a restaurant gets, not how it was produced behind the scenes.
Naming a vendor behind a live Dohos capability anywhere on this site — even where a real working relationship exists. Search for one; the search comes up empty by design.
PROVEN AT SUBPROCESSORS →Anything the system isn't certain about goes to a person. A wrong guess costs more than a short delay.
Silently picking the most likely reading of an ambiguous request and hoping — a quiet guess reads confident in the moment and turns up wrong on a ticket nobody checks until it's late.
PROVEN AT MENU →No invented metrics, ever — internally or in front of a customer.
A plausible-sounding number standing in for one nobody measured. Where no verified figure exists, this site prints the gap — a placeholder glyph or a plain sentence — never a guess.
PROVEN AT CALLS →What's spoken on a call and what prints on a ticket come from the same record, with no second copy to drift.
A second copy of a price or an 86 that can quietly disagree — three surfaces touch one record, never three records that need reconciling by hand.
PROVEN AT MENU →Each one fails visibly the moment it isn't kept.
That's the actual bar for what belongs on this page: not whether a sentence sounds good, but whether its failure would be specific, visible, and checkable.
The clearest example of a practice with a real cost: escalating rather than guessing means a caller sometimes waits and a manager sometimes gets interrupted for something that turns out to be simple. That tradeoff is accepted deliberately. Between a visible delay and an invisible wrong guess, the delay is the honest cost — the alternative is a system that's confidently wrong some real share of the time, in a way nobody catches until a ticket arrives with the wrong item on it. The allergy case is where this matters most, and menu states that limit directly: an AI response can't guarantee an item is safe for a specific allergy, so it doesn't pretend to.
The obvious objections, taken seriously.
Aren't these just things every company says?
Said, maybe — plenty of companies publish something like this. The difference is that each practice here links to a mechanism you can independently check. “Escalate rather than guess” isn't proven by the sentence — it's proven by the resolver behavior and the allergy-handling limit described on menu. Read those, and the claim either holds or it doesn't. That's a materially different test than whether a phrase sounds sincere.
Why isn't the list longer?
Because most of what could go on a longer list wouldn't clear the bar — a real, checkable mechanism behind it. “We put customers first” has no failure condition, which is exactly why it reads as a slogan. This page stays narrow on purpose, to keep every entry provable.
Does this page apply to itself?
It has to. Every mechanism named above links to the page that documents it precisely so none of these sentences needs to be taken on faith — and the no-invented-figures practice is checkable against this exact page, which prints no number anywhere.