VETRA

Vetra for Restaurants & Cafés

Reviews and no-shows.
Vetra protects the table.

A table held for a guest who never arrives, and a bad night that reaches your reviews before it reaches you — those are the two quiet leaks in an independent dining room. Vetra answers the phone with your real menu and hours, captures the enquiry as a lead, asks every guest how the visit went, and keeps the guest list that tells you who each one was. The same engine runs the café half of this page, asking the counter version of the same questions.

Independent restaurants and cafés, one to six locations. Square, Toast — or a register with no API at all.

Already in your POS Nothing has asked it
Every cover and order, timestampedWhich hours are actually dead
Item mix, and your costsWhat earns, versus what merely sells
Who came in, and how oftenWhich guests and regulars stopped
Your menu, hours and 86 listWhat the phone tells a caller

The no-show

A table held for somebody who never arrives.

It starts as a phone call at the worst possible moment — mid-service, hands full, nobody free to pick up. The call goes to voicemail, the party books somewhere else, or they book with you and you never hear from them again. Vetra takes that call and gives you back the one thing voicemail does not: a name, a number and what they wanted.

Enquiry captured Awaiting your confirmation
Party of six, Friday 7:30pm
Caller left a name and a mobile · asked about the tasting menu
You confirm
"Are you open Monday?"
Answered from your hours · no lead created, nothing for you to do
Handled

Sample calls — an invented restaurant. A real call reads your real hours and menu.

It never says "confirmed".

Vetra holds no table inventory, so the agent takes the request and hands it to you. In every restaurant conversation it is held to the same instruction — never tell a guest a booking is confirmed — because the seat is a commitment only you can make.

The 86 list travels with it.

What the kitchen took off tonight is what the phone, the widget and the dashboard all say. A caller is never offered a dish you cannot serve.

You get the summary, not the voicemail.

Who called, what they asked for, and the transcript. It lands in notifications and as a text, so the follow-up happens the same night instead of the next morning.

The review

You should not learn about the night from a screenshot.

A guest has a poor night, says nothing, and writes it down three days later where everybody can read it. The gap between the visit and the review is the only place anyone can do anything about it — so Vetra sends a short text into that gap and asks how it went.

The check-inBoth modes

One text after the visit, in your name. A poor answer opens a private make-it-right conversation and alerts you the same day, so you hear it from the guest rather than from a stranger's phone.

The same link for everyoneDesign law

Whichever way a guest answers, the same Google review link goes out. Nobody is sorted by what they said and nobody is skipped for it. That is not a setting you could switch — it is written into the engine, permanently, and it is the one thing here we would not change for a customer who asked.

The room, read back to you

Covers, dayparts, and the guests behind them.

The restaurant dashboard is the same engine as the café one with the vocabulary changed — covers rather than orders, average check rather than ticket, dayparts rather than the morning rush. Ask it in the words you would use out loud and it answers off your own sales.

Your guest list, in four bandsBoth modes

Regulars, the ones who come back sometimes, the ones drifting away, and the first-timers — built from what they actually bought, not from anyone filling in a form.

The dayparts that sagBoth modes

Which service is quiet, on your own trading pattern rather than an industry chart — and the busy ones too, so nothing gets discounted into a full room.

One kitchen, several roomsMulti-location

A group view across your locations, and each room on its own page underneath it. Built for an operator with a handful of sites, not a franchise system.

The café half — the counter

Your POS has been collecting the answer for years.

And it has never once told you anything. Every order is in there, timestamped. Your item costs are in there. So is every regular, and the week each one quietly stopped coming. Vetra reads what you already pay to collect — and tells you what it means. Everything from here down is that, in full, for an owner who still works shifts and runs one to six shops.

The quiet number

Nobody announces they are leaving. They just stop.

There's no cancellation, no complaint, no exit survey. A regular comes in twice a week for a year, then doesn't — and the first time you notice is months later, if at all. It's already in your POS. Nothing has ever asked it.

Lapsed regulars 3 people
Dana R.
41 visits · last in 22 days ago · usually Tue 8am
Lapsed
Marcus T.
9 visits · last in 17 days ago · usually Fri 3pm
Lapsed
Priya N.
4 visits · last in 15 days ago · usually Mon 7am
Lapsed

Sample data — invented people at an invented shop. No real customer appears here.

Dana is worth reaching. Priya is a drop-in. Knowing which is which is the difference between a gesture and a giveaway.

One POS connection

One connection. Six things it can finally tell you.

It all comes off sales you are already recording. Open any card and use it — these are the real components, not screenshots of them.

Ten tools sit over your live sales — slow windows, top and bottom items, per-item margin, lapsed regulars, campaign redemption, live 86 status. You type the question in the words you would use out loud.

Answers computed in code, never estimated by a model Every figure traces to a query you can open A question it cannot answer gets told so, plainly Same tools the voice agent uses, so both agree
Ask Meridian Coffee10 tools

Pick a question below, or type your own.

Sample data — an invented shop. Nothing real appears on this page.

Volume and margin are different questions, and a POS report only ever answers the first. Put your costs in once and the menu sorts itself into four boxes — the useful one being the plow horse: high volume, low margin, and almost always the drink you are proudest of.

Margin per size, not a blended average per item Stars, plow horses, puzzles and dogs, on your own numbers Costs you enter once and revise when a supplier moves Knowing it is a plow horse does not mean cutting it
Menu quadrant · last 30 daysNothing selected
Margin Volume
Puzzle Star Dog Plow horse

Sample data — an invented shop's menu and costs.

Click the dot bottom-right. That is the plow horse — the one selling hardest and earning least. What you do about it is a judgement call, and the point is that right now nothing tells you it is happening at all.

A real scraper reads nearby shops' published menus twice a month, normalises the delivery markup out, and ranks where you sit. It produces a recommendation with its reasoning attached — and then it stops, because it is your menu.

Live sources, not a stale price database Delivery-app markup normalised out before comparing Refreshed on the 1st and the 15th We never write a price to your POS. Not once, not ever.
12oz latte · within half a milerefreshed 1 Aug
Fourth & Bloom$5.750.3 mi
The Wheelhouse$5.500.2 mi
Meridian Coffee — you$4.75
Ridgeline Roasters$5.250.4 mi
Corner Cup$4.500.5 mi

Sample data — invented shops at invented prices.

RecommendationAdvisory

You are 75 cents under the local median. Four of five shops within half a mile are above you, and the two nearest are at $5.50 and $5.75. Moving to $5.25 would put you mid-band rather than cheapest.

Whether that is right depends on things we cannot see — who your regulars are, what you want to be known for, and whether being the cheapest on the block is the point. So it is a recommendation, and it stays one.

Every owner knows the afternoon sags. Almost none can say which afternoon, by how much, or whether it is the same one each week. That is a query, not a hunch — and it is the difference between a blanket promo and one pointed somewhere.

Your own thirty days, not an industry daypart chart Peak hours too, so you never discount into a rush An offer sized against margin, not pulled from the air You approve every send. Nothing goes out on its own.
Sales by hour · last 30 daysPick a window
Quiet Busy

Sample data — an invented shop's trading pattern.

Click the quietest square you can find. Tuesday afternoon is usually the one. Then look at what an offer aimed only at that window looks like, versus the same money sprayed across the week.

The definition is deliberately narrow: someone who came at least twice and has not been in for fourteen days or more. Not a marketing segment — a short list of people whose absence is a change in behaviour rather than a coincidence.

Built from purchases, not from anyone filling in a form Their own pattern, so you can see what changed Short enough to actually read on a Tuesday morning A reason to reach out, not an automated send
Lapsed regulars3 people

Sample data — invented people at an invented shop. No real customer appears here.

Open one. The useful part is not that they are gone — it is what their pattern was, because a 41-visit Tuesday regular and a four-visit drop-in are not the same loss and should not get the same message.

The voice agent is not a separate product with its own facts. It calls the same tools the dashboard does, so if it says an item is sold out, that is because the 86 list says so — and the answer it gives a caller is the answer you would have given.

Answers on the first ring, through the grinder Every answer traced to a field you set, never invented Catering and large orders captured, not lost to voicemail Bundled in. It is not a hundred-dollar add-on.
Incoming callIdle

Take a call below to watch it run.

Sample call — an invented caller. A real call reads your real menu and hours.

Pick a caller. Watch the third one — she is on the lapsed list in card 05, and taking her call updates that card while you are looking at it. Same system, not six things sold in a bundle.

Before you ask whether it works with your setup

Square, Toast, or a register from 2009.

Same engines either way. Only where the numbers come from changes — and if they come from the second path, we tell you exactly what that costs you in precision.

If you run Square or Toast

Connect it, and it reads itself.

  • Sales, menu and customers sync on their own
  • Square connects with one OAuth click
  • An approved campaign promo is written as a real POS discount
  • Items you 86 flow through to the AI and the phone

The promo is the one thing we write to your POS, and it only moves after you approve it. Prices we never touch.

If you run an old register

Then we sit on top of it.

  • Orders through voice, widget and catering settle to your bank
  • Menu built from a photo or a paste of the one you have
  • Receipt capture reads paper into the same tables
  • An 86 you set by hand persists everywhere

And the honest part: we do not see every cash sale. We see what comes through us. The analytics are directional, not your books — and we would rather say that now than have you find it in month two.

The everyday win

The three screens owners actually open.

Insights answers the question you thought of. Today surfaces the one you did not. The composer turns either into something you can send.

Untaggedruns today In buildon the queue, not shipped yet — we say so rather than wait
AI Insights — plain-English questionsBoth modes

Ask "what are my slowest hours?" and get back a chart, a ranked list or a written answer — with every figure tracing to a real calculation rather than an estimate.

Today — the daily operating briefBoth modes

Yesterday's revenue, lifetime repeat rate, average rating and a ranked action queue. Each card deep-links to the tool that acts on it with the message already written.

Campaign composerBoth modes

Type a goal — "dead Tuesday, send 10% off" — or pick a playbook. It counts your exact audience and checks the discount against a 20% margin floor, warning you before you cut past it. On approval the promo is written to your register as a real discount.

Acts, not just answersBoth modes

It proposes setting changes — catering lead time, say — as a before-and-after with Confirm and Cancel. Applied atomically, and written to the audit trail either way.

Make more per cup

The pricing cockpit.

Competitor scraping, your costs, the quadrant, and a tuner that does the what-if maths live. Advisory from end to end — the last step is always yours.

Price tuner and what-ifBoth modes

Market low, average and high as a band with a slider. Live margin, revenue and profit deltas at any price, with a 20% margin-floor warning before you overshoot.

One-press price changeIn build

On Square, approve a recommendation and it writes to your catalog. On Toast, we hand you the recommendation to apply. Either way you approve first — nothing moves on its own.

Closed-loop proofBoth modes

After a price change the dashboard tracks predicted against actual impact, with a lifetime profit-added total and a running accuracy figure. You see whether the calls were right, because a number you cannot check is not proof.

COGS and auto-recipeIn build

Paste your menu and it estimates the ingredient recipe and cost behind each drink, so everything downstream runs on profit rather than revenue. You correct anything it gets wrong, and it remembers.

Vetra does not change your prices. Not with a button, not overnight, not with permission given in advance. You get a recommendation and the reasoning behind it. The change happens in your POS, by you.

What the research actually found

Most discounts reach people who were already coming.

Punch cards do work — and the research on what they actually do isn't the research the vendors quote. A discount is only an investment if it's aimed. Broadcast, it is margin handed to people who were already walking in.

Kivetz, Urminsky & Zheng · 2006

A real multi-location campus café, buy-ten-get-one, 949 completed cards across roughly 10,000 purchases. Regulars moved from about nine to eleven purchases a month.

The part the vendors never quote is in the same paper: purchase rates slow down again once the reward is earned. It isn't habit formation. It resets.

Journal of Marketing Research 43(1)

Leenheer et al. · 2007

Correcting for self-selection shrinks the measured share-of-wallet effect roughly sevenfold.

Which is the shape of every headline number in this category. Restaurants that bought marketing sell more. Customers who enrolled spend more. Both compare people who opted in against everybody else, and call the gap a result.

Intl. Journal of Research in Marketing 24(1)

Sharp & Sharp · 1997

Loyalty schemes are "quite defensive in nature, encouraging buying by existing buyers."

In their data the scheme's share still fell from 32% to 30%, while a rival running plain price promotions went from 21% to 28%.

Intl. Journal of Research in Marketing 14(5)

"A program that appears to drive visits while eroding 8-10 margin points through excessive discounts isn't driving ROI — it's subsidizing behavior you may have gotten anyway."
— Thanx, a loyalty vendor, on their own pricing page. thanx.com/pricing

So the useful question is not how do I run a loyalty program. It is which customer, and which hour — because that is what turns a discount from a giveaway into an investment. Everything below is a way of answering it, off data your shop already has.

Keep them coming back

You decide. We aim it.

Messaging is enabled per shop and the campaign is yours — you write it, you confirm who it goes to, you press send. What Vetra brings is the aim.

Win-back and drip SMSBoth modes

Triggered texts for slow days, lapsed customers and birthdays, written per customer. Daily send caps and opt-out suppression are enforced in the send gate, not left to config.

Hybrid loyalty loopNo-POS / Hybrid

Milestone lifecycle texts off order data — welcome on the first, win-back at N days lapsed, a reward every Nth visit. Frequency-capped and quiet-hours aware. Points and visits live in Vetra, not inside Square or Toast — if you already run theirs, this sits alongside it rather than replacing it.

The check-inBoth modes

A short text asking how the visit went. An unhappy answer opens a private make-it-right conversation and alerts you the same day, so you hear it from them rather than from a screenshot.

Lapsed-VIP finderBoth modes

Customers with two or more visits who have not returned in fourteen days or more, with their own pattern attached, and one click to launch the win-back. Card 05 above is this. The revenue-at-stake figure beside each name is in build — the list is live now.

Compliance, stated precisely.

Messaging is enabled per shop. "Reply STOP to opt out" is appended by the send gate automatically, and opt-outs suppress every engine. Frequency caps and quiet-hours checks are enforced in the loyalty loop. Texts go from a single TCR-registered toll-free number, and consent is captured before anyone is messaged.

The front desk that never sleeps

Voice, widget and catering. Bundled.

The same front desk for both rooms — the restaurant's reservation enquiry and the café's "are you open" land in the same place. Not a hundred-dollar add-on at the bottom of an order form. It is in the price at every tier.

Voice AIBundled

Answers calls, reads your menu and hours, captures a name and number as a lead. Transcript and recording land in notifications. Never offers an item you have 86'd.

Website widgetBundled

Customer-facing chat on your own site — menu, hours, lead capture. The same brain as the phone, so the two never disagree. One snippet.

Catering funnel and CRMBoth modes

The high-ticket work most café and restaurant tools ignore. Public intake form, validated against your own rules, into a new / replied / won / lost pipeline. The phone honours those same rules.

Channel ROIBoth modes

Leads and converted revenue attributed by channel — voice, widget, catering, missed call — with a period-over-period trend, so you can see which one actually pays.

The QR signupBoth modes

A code provisioned per location for the counter or the table. Whoever scans it links themselves to their own order history — no app, no punch card, no friction at the till.

Receipt-to-profile readerIn build

Photographs of receipts read into purchase history — what was ordered, how often, which day — so the regular list builds itself without manual entry. The same reader runs on bar slips; the café and dining-room path is on the queue.

It gets better the longer it runs

Month six knows things month one could not.

Not a slogan — a consequence. Every one of these needs accumulated trading to work at all, and the dashboard tells you which ones are not there yet rather than pretending.

Maturity stages, shown to youStage 0 → 3

The dashboard shows what the system can and cannot yet do, and what unlocks next — "100 orders and it can recognise the usual." Visible progress rather than a black box.

Demand forecastWeather-aware

Tomorrow's revenue and orders, weather-adjusted, carrying a live accuracy figure that improves as data accumulates — including learning that weather does not move your sales.

Network benchmarksCross-room

You against the average Vetra room of your own kind on list size and daily orders. It gets more useful as more shops join, which we would rather say than pretend it is already dense.

Settings write-logFull audit

Every change logged with an AI-versus-owner tag, the before and after values, and a timestamp. If something moved, you can find out what moved it.

Published, because nobody else in this category publishes

One price, and it's on the page.

Toast Loyalty, Thanx, Punchh, Paytronix, Ovation — every one of them gates the price behind a demo. Here is ours: $249 a month, per location, 400 voice minutes included, $0.20 a minute past them.

Vetra for Restaurants & Cafés

$249 / month

400 voice minutes included every month $0.20 per minute past the included 400 Every card on this page, the phone, one location Runs on any POS — Square, Toast, or a register with no API

Minutes over the included 400 are metered and land on the following month's invoice at $0.20 each. Multi-location is coming and is deliberately unpriced until it exists.

Talk to Avi first.

Twenty minutes on your room, your menu and your POS — he built it and he is the person who will answer. Or skip it and start below; two fields and you land on Stripe.

Stripe sends the receipt here, and it becomes your login.

Payment is handled by Stripe. Vetra never sees or holds your card or bank details.

Talk it through

Twenty minutes, and we build to your room.

The price is above and the checkout is real, so this is not a gate — it is for the owner who would rather watch it run against their own menu first. You get the founder's attention and a setup that bends to how your room actually runs; we get somebody telling us what is wrong with it.

  • Connect Square or Toast, or start with no POS at all
  • Month to month — no hardware contract to begin

Book the demo, or send this instead.

Not a sales team — there isn't one. He built it, and he is the person who will answer. Independent restaurants and cafés, one to six locations.

No POS is fine — that path is built. "Not sure" is fine too.