Pizza Checker · Algovectra

Why it matters

One bad pizza costs more than one pizza.

Customers forget the good nights. They remember the burnt one, and they act on it.

For a pizzeria

You lose a regular.

Unhappy delivery customers rarely complain. They just stop ordering.
They can't send it back to the kitchen, so next Friday they order somewhere else and you never find out why.
You're judged by your worst shift.
Most nights your pizza is great. Then one hectic rush or one new hire at the oven sends out a bad one, and a regular who ordered every week is gone.
Your rating is your shop window.
A one-star review with a photo of a burnt crust greets every new customer who finds you on a delivery app.
Your standard leaves when you do.
You can't be at the oven every shift. Pizza Checker holds the kitchen to your standard while you're away, and you can look at every pizza wherever you are.

For a chain

The whole brand pays for it.

Customers don't tell branches apart.
Bad news travels faster than good pizza. When one kitchen slips, people say the whole chain has gone downhill.
Consistency is what people buy from a chain.
They choose your brand because they know exactly what they'll get. Without a measured standard, you're just pizzerias sharing a logo.
Audits see a good day.
Inspections and mystery shoppers check a handful of pizzas. Quality slips at 8 pm on a Friday, and it only shows in the ratings weeks later.
What isn't measured can't be rewarded.
If branches are judged only on speed and food cost, quality is the first thing to give. Measure it per branch and you can set fair KPIs and reward the teams that keep the standard.

The packing station is the last moment a bad pizza can still be fixed. That's where Pizza Checker works.

How it works

Each pizza is checked without extra work on the line.

The demo below uses real production photos and our models' real output. Every order goes through the same checks.

The same pizza isolated from its box — Photo taken as the order is packed

Doneness map of the crust, mostly blue, which means underbaked — Photo taken as the order is packed

Complaints

Answer complaints with a photo of what you sent.

Every pizza is photographed and checked when it's packed, and the result is saved with the order. Connected to your complaint system, Pizza Checker takes the customer's photo, checks that it shows the same pizza, and then checks the claim itself against what the camera saw at packing, for example whether a topping was really missing or the crust really burnt. The complaint is refunded or declined by your rules, and the reply is written for you.

A photo and every check result saved with each order
The customer's photo matched to the photo from packing
The claim itself checked, such as whether an ingredient was really missing
Refund or decline by your rules, with the reply written for you
Complaint statistics that show exactly what was wrong, down to the missing topping

How a complaint is handled

Step 1 of 9 · Order

A customer orders a Rustica.

Schematic of complaint handling connected to your complaint system. The two example orders are real production captures with their real check results, the topping model's real olive detections, and the doneness model's real map. The other rows come from the product's order history. The customer's photo is made from the same capture, shown as a phone shot at home. Complaints, replies, and statistics are illustrative.

For chains and franchises

One measured standard for every branch.

With the same check on every pizza, quality becomes a standard you can manage across the whole network. It also gives you a fair basis for rewarding the branches that keep to it.

Network quality · this week

Illustrative data

Illustrative ranking of branches by pizza quality
Branch	Pass rate	Burnt	Portions	Trend
1Riverside	98.6%	0.4%	On spec	Improving
2Old Town	97.9%	0.8%	On spec	Improving
3Station Square	97.2%	1.1%	On spec	Stable
4Airport	95.4%	2.3%	Over-topped	Declining
5North Park	91.8%	4.6%	Under-topped	Declining

Network standard: 96% pass rate1 branch needs attention

All branches on one scale

Pass rate, burnt rate, and portion accuracy compared across branches, shifts, and products.

Fair KPIs, fair rewards

Bonuses follow measured quality instead of impressions. Good teams get the credit, and the rest get specific feedback on what to improve.

Early warning

You get an alert when a branch starts to slip, well before it shows up in the ratings.

Launches done right everywhere

See from day one which branches make the new pizza right.

Margin you can see

See which branches put on too much topping and which ones skimp.

One admin for the network

Users, roles, branches, and cameras managed centrally, with data ready for your BI.

Satisfied customer

vosime.cz

Vosíme runs Pizza Checker in production.

A pizza delivery chain with 40+ branches making 220,000 pizzas a month. Vosíme is a satisfied customer and is rolling Pizza Checker out across its network.

branches in the chain: 40+
pizzas a month across the chain: 220,000
pizzas evaluated by Pizza Checker: 1M

Chain figures describe Vosíme as a whole, not current deployed coverage of Pizza Checker. Figures as of September 2026.

See the Vosíme implementation

The technology

Made for checking pizza and nothing else.

General-purpose AI knows a little about everything. Pizza Checker only knows pizza. We build the whole stack ourselves, from the camera to the specialised models, which have already seen a million real pizzas.

Built in-house, layer by layer

Our camera
Designed and built in-house so every photo is taken the same way: 4K images, locked exposure and white balance, and motorised focus.
Our models
Model architectures and algorithms designed for checking pizza and trained on real production data. We don't hand a general model a photo and ask it to “look at this pizza”.
Our cloud
We run, monitor, and update everything. All you see are the results.

Pizza Checker compared with general-purpose AI
Specialist vs. generalist	Pizza Checker	General AIGeneral-purpose AI
Built for	Pizza inspection	Everything
Hard-to-see ingredientslike a pickle under melted cheese	Spotted	Easily missed
Full analysis of a 4K photo	Under 1 second	Several seconds
Cost per pizza	A fraction	Many times more
Same photo, same verdict	Always	Not guaranteed

Pizza Checker timing runs from photo upload to verdict at full 4K. The general-purpose column describes typical large multimodal AI models, not a specific product.

Our engineers build the models.

Pizza Checker's models are designed, trained, and improved in-house by our own computer-vision engineers, who work on pizza analysis every day. We don't resell someone else's AI. When your menu or your kitchen needs something new, the people who built the models do the work.

Meet the team

Our camera

Our own camera, built for the pass.

We design and build the camera ourselves. That's how we fit a 4K autofocus camera, its own computer, and its cooling into a sealed aluminium body of 157 × 65 × 35 mm. It copes with a kitchen's heat, steam, and humidity, and it mounts flat under the shelf, so it takes up no space on the counter. Because the hardware is ours, we aren't held back by the limits of an off-the-shelf camera and can keep extending what a camera at the packing station can do.

Sensor: 4K stills, 4056 × 3040 px
Focus: Motorised autofocus, adjustable from the browser
Exposure and white balance: Set remotely, then locked so every photo matches
Power and data: One PoE+ cable (IEEE 802.3at)
Network: Plug and play, needs only internet access (no network setup or VPN)
Size: 157 × 65 × 35 mm, mounted flat under the shelf
Body: Sealed aluminium, no open ports
Made for hot kitchens: Up to 50 °C ambient, resistant to steam and humidity
On board: Its own computer takes and uploads every photo
Updates: Regular over-the-air firmware updates

Rendered from the camera's CAD model, shown in the production finish.

Installation

Installed in three steps.

The camera is plug and play: mount it, plug in one cable, and focus it from a browser. You don't need a power socket at the shelf or any network setup, and there's nothing to adjust by hand on the camera.

Drawing not to scale. The camera is rendered from its CAD model; the pizza is a real production capture.

Remote management

Adjust it from anywhere, with no trip to the branch.

All cameras are managed in our web admin. From one place you can focus any camera in the network, set its exposure and white balance, take a test photo, or restart it. A blurry photo gets fixed from the office instead of with a drive out to the kitchen, and new firmware arrives over the air.

Autofocus, focus on a chosen spot, or by hand
Exposure and white balance
Test photo and restart
Firmware updated over the air
The status of every camera in the network

Try it: run autofocus, drag the lens off focus, or change the exposure.

Live photo from the camera: a boxed pizza on the packing counter, seen from above — Illustrative view of the camera admin, redrawn in this page's design. The photo is a real production capture and the exposure readout comes from its metadata; branch names and other values are examples. In the product, each change is applied on the camera and confirmed with a new photo.

Open API

Run your business on data from every order.

Pizza Checker has an open REST API documented in OpenAPI, so it connects to whatever you already use. Everything it records about your orders is yours: the photo of each pizza, the result of every check, what was missing or burnt, the pizza type, the branch, and when it left the oven and the packing station.

Load it into SAP, a data warehouse, your BI dashboards, or accounting, and quality becomes a number you can plan with, next to sales, food cost, and staffing.

Ask about API access

One checked pizza from the API

OpenAPI 3.1 · JSON

{  "name": "Rustica",  "capture_at": "2026-08-27T17:45:22Z",  "evaluation": {    "metadata": {      "overall_evaluation": { "status": "OK" },      "components_evaluation": {        "base_components": {          "expected": ["salami", "mushrooms", "onion", "olives", "bacon bits"],          "found": ["salami", "mushrooms", "onion", "olives", "bacon bits"],          "missing": [],          "status": "OK"        }      }    }  }}

A real check of one pizza, shortened, with ingredient names translated from Czech. The full record also links to the photo and the check layers, and carries the branch, prices, and times from the kitchen.

SAP and other ERP
Tie quality to food cost, stock, and the results of each branch.
Data warehouse
Keep the full history of checks and photos next to your sales data.
BI and analytics
Build reports in the BI tool your analysts already use.
Accounting
Match refunds and write-offs to the orders and photos behind them.
POS and ordering
BiteBerry is connected out of the box. Other systems connect through the API.
Your own apps
Use the photos and results in support, staff training, or franchise audits.

What the data tells you

Questions you used to answer on gut feeling now have numbers behind them.

1Which pizzas most often leave the kitchen with a topping missing
2Which branch, and which shift, burns the most crusts
3What bad pizzas and refunds cost you each month
4How long orders take from the oven to the packing station

Getting started

Start small and pay per pizza checked.

Pricing

Pizza Checker is a service.

AI vision isn't something you install once and forget. It behaves more like a living organism that grows with your kitchen: you add pizzas, change recipes, bring in new ingredients, or want pizzas judged a little differently, and Pizza Checker changes with you.

With boxed software you pay a large sum before it checks a single pizza, and every later change costs extra. With Pizza Checker you pay only for the pizzas that get checked. Support is included, and within your plan you can adjust the evaluation however you like.

There are no servers to buy and no AI infrastructure to run, so it pays off even for a single pizzeria.

What you pay, month by month

First two years

Nothing checked yet
Checking pizzas
Change included in your plan

Illustrative, not to scale. Your price depends on your volume and setup, so ask us for a quote.

Per-pizza pricing
You pay only for relevant evaluations of your pizzas.
100 hours of support a year
Time with our team for questions and requests, included in the price.
Changes within your plan
Add a pizza, change a recipe, bring in a new ingredient, or adjust the evaluation settings whenever you like.
Operation included
Hosting, running, and monitoring in our cloud are part of the price.
Costs follow your volume
The price scales from one pizzeria to a whole network.
Camera options on request
We match them to your volume and setup.

Get a quote for your volume

Live on your menu within a month.

Pizza Checker is a ready product, so there's no development project to pay for. It already recognises 30+ ingredient types, and we only teach it the ingredients and sauces specific to your menu.

Integrations: Integrated with BiteBerry; we connect to other POS, ordering, and delivery-management systems.
What you need: A spot above the packing station with even lighting, and one network cable with internet access. There's no network setup and no VPN.

About the camera hardware

1
Install the camera
Mount it above the packing station. For the first weeks, it collects images of your real production.
2
We teach it your menu
We build an annotation plan and teach the models your specific ingredients and sauces.
3
Go live
Checks run on every order from then on, typically starting within a month of installation.

FAQ

Questions operators ask.

Does it work for a single pizzeria?

Yes. You pay per pizza checked and there's no big upfront investment, so a single branch can start small. Chains usually start the same way, with one pilot branch.

Do we need BiteBerry?

No. BiteBerry is integrated out of the box, and we connect Pizza Checker to other POS, ordering, and delivery-management systems.

Does it slow down the line?

There's no extra step. The photo is taken when staff mark the order as packed, and the full analysis takes under a second once the photo reaches our cloud. An LED strip shows staff when to keep hands out of the frame.

Does it watch my staff?

It's pointed at the pizza. The camera looks straight down at the packing table and takes a photo when the order is packed.

Does someone have to come to the branch to adjust the camera?

No. The camera has motorised autofocus, and its focus, exposure, and white balance are adjusted remotely from our web admin. On site it needs just one cable for power and data.

What does the camera need from our network?

Just internet access. The camera is plug and play: it needs no network setup or VPN, connects to our cloud by itself, and gets its firmware updates over the air.

How fast does our internet connection need to be?

We recommend an upload speed of at least 50 Mbps; download speed doesn't matter. Pizza Checker also runs on a slower upload, but each check then takes longer, because every 4K photo has to reach our cloud before it's evaluated.

Why not just use a general AI model?

General models are made to handle anything. Our engineers built ours for pizza only, so they spot hard-to-see ingredients, finish the full analysis in under a second, cost a fraction to run, and always give the same verdict for the same photo.

What happens when our menu changes?

We adjust Pizza Checker to match, within your plan. New pizzas, recipe changes, new ingredients and sauces, and different evaluation settings are taught the same way as during onboarding.

How much does it cost?

Pizza Checker is a service, and you pay per pizza checked. The price includes cloud operation, 100 hours of support a year, and changes to the evaluation within your plan. Camera options and the exact price depend on your volume, so ask us for a quote.

* The film at the top of the page. Animated film. The kitchen scenes were generated with AI from photos of a Vosíme branch; the camera is a 3D render of its CAD model in the production finish, and the map is a 3D render. The pizzas are real production captures with the models' real detections and verdicts; the branch network, its scores, and the charts are illustrative. Aerial imagery: Ortofoto © ČÚZK, CC BY 4.0; Sentinel-2 cloudless - s2maps.eu by EOX IT Services GmbH (contains modified Copernicus Sentinel data 2016), CC BY 4.0.

Catch the bad pizza before your customer does.