Privacy policy
Version 2026-10-01, in effect from 1 October 2026. Last updated 1 October 2026.
Counter Admin turns a paper Menu into a digital one. This policy says what the service stores while it does that, how long it keeps it, which other companies see it, and how to have it deleted. It describes what the product does today, not what it may do later.
Who is responsible
- Registered name
- Counter Admin
The operator is the data controller for Restaurant accounts and for anyone who scans a Menu without an account. For a diner who opens a published Menu, the Restaurant is the controller of its own Menu content and this service is its processor.
The service is run from the United Kingdom and sold to Restaurants anywhere. UK GDPR is the baseline it is built to. Where the EU GDPR applies to a Restaurant in the European Economic Area, the same rights apply and the Information Commissioner's Office (ICO) in the United Kingdom is not the only authority you may complain to: you may also go to the supervisory authority where you live or work. Nothing here reduces a protection your own country's law gives you.
Privacy questions go to contact@counteradmin.net.
What is collected
Account data
Your email address, your display name, and the Organization and Restaurant details you enter. The address and the password live with the authentication provider, not in this service's own database, and no password is ever stored or readable here. If you sign in with Google, what reaches the service is your address, the name on the Google account and the fact that Google had already confirmed the address. No profile picture is read or stored.
People you invite
An invitation stores the address you typed, the role you chose, who sent it and when it expires, until it is accepted or revoked. The service also keeps a log of which messages it sent to which account and whether they were delivered. That log records the account and the kind of message, never the address or the text.
Menu content
Everything you put on a Menu: Item names, descriptions, prices, Categories, Offers, opening hours, contact details, translations, dietary and allergen markers, Item photos, and your logo. This is business content, and the part you publish is public by design.
Uploaded Menu documents
When you import a Menu you upload photographs or a PDF of it. Those files go to a private storage bucket, readable only through a request that has already proved it belongs to your Restaurant, and separate from the bucket that serves published images. Each page is sent to OpenAI as an image or a PDF so it can be read back as text. If your paper Menu happens to carry a person's name or a photograph of a person, that reaches the provider along with the rest of the page, so do not upload a document carrying personal information you do not want processed.
What comes back is a draft: the Menu text the model read, stored with the import so you can check it line by line. Nothing in it reaches a live Menu until you confirm it.
Billing data
Subscription payments run through Stripe. Card numbers never reach this service, and neither does a card brand, the last four digits, a billing address or a tax number. What is stored is the Stripe customer and subscription identifiers, which plan you are on, the status, the current billing period, whether you have cancelled, and the identifiers of the events already processed so none is applied twice.
Service email
Password resets, confirmations, invitations and notices the service has to send go through Resend, which sees your address and the message. There is no marketing email: the product has no newsletter and no code to send one.
Technical records
Each model-assisted run records only metadata: which provider and model ran, the prompt version, a request identifier, how long it took, token counts, and a failure code if it failed. The uploaded bytes, the prompt and the raw provider response are never written to those records.
When a request fails, the error is reported to the monitoring service as the error message, the route pattern and an operation name, with credentials, tokens, connection strings and email addresses stripped out first. Nothing else about the request is attached. These reports are made on the server; nothing reaches the monitoring service from your browser.
The service also counts requests per short window to stop one client flooding an endpoint. Those counters hold a count and an expiry, and they are overwritten as the window moves on.
What is not collected
- Nothing identifying from diners. A published Menu loads no third-party analytics, no tag manager and no third-party script, sends a diner no cookie, and serves its fonts from this service rather than fetching them from another company. Opening a Restaurant's Menu on a phone is never profiled or tracked. It does add one to a daily total for that Restaurant, which is described under Menu visits below and holds nothing about the person who opened it.
- No session recording. The product has no session replay, no heatmap and no keystroke capture. None of it is installed anywhere.
- No card data. Stripe holds it.
- No advertising. Your data is never sold, rented or shared with an advertising network.
- No model training. Every request to OpenAI is sent with storage switched off, so the provider does not keep your document or the text read from it against this service's account, and OpenAI does not train its models on data sent through its interface. The provider may still hold a request briefly for its own abuse monitoring, under its own terms.
Reports you send us
A button on the product screens lets you report a problem or ask for a feature. It is never shown to a diner on a published Menu. Nothing is sent until you press send.
What a report carries
Your note, a picture of the screen you were on with whatever you drew on it, and a close crop of the part you drew around. You can switch the picture off before you send and send the note on its own.
Attached without you typing anything: the page you were on and the identifiers in its address, the release the screen came from, and technical facts about the browser and the window, which are the address it was served from, the browser and platform it reports itself as, your language, your time zone, the size of the window and of the screen, how far down the page you had scrolled, whether the device is touch, whether you are in dark mode, whether you have asked for reduced motion, and whether the browser believed it was online. Two of your own choices go with it: whether the report button is switched on, and whether you have allowed workspace analytics.
It also carries what went wrong just before you pressed send: at most the last 20 errors the browser saw and the last 20 requests that failed or took more than three seconds. Web addresses in that list are recorded without their query strings, because that is where identifiers and one-time links sit.
If you are signed in, the report says who sent it, using the name and address your account already carries, and adds the Restaurant and Organization the page belonged to, your role, whether you were in read-only mode, your Plan and its status, and the Entitlements that produced the screen. Nothing new is collected about you to do that.
If you are not signed in, there is no account to name, and the operator still has to tell one stranger's reports from another's. So the report carries a two-word nickname worked out from the random identifier the feedback tool had already given this browser, together with an approximate country and city. That nickname belongs to a browser rather than to a person: another device, another browser or clearing your site data produces a different one, and nothing measured about you goes into it. The country and city are worked out by Cloudflare at the edge, from the connection that is already reaching it. The service never reads, stores or sends your address, no other company is asked where you are, and the place is kept only as part of that one line.
The picture is the sensitive part
It is a photograph of that screen at that moment, so it carries everything the page was showing, not only the thing you are reporting. If a colleague's name and address, an invitation you had typed, or anything else about another person was on screen, it is in the picture and it reaches us. Look at what else is in the frame before you send. If you would rather not send a picture at all, switch it off, or write to contact@counteradmin.net instead and send nothing but your words.
Where it goes and who reads it
Reports go to a separate Google Firebase project the operator runs for collecting feedback, not to this service's own database. That project is shared with the operator's other apps: every report is tagged with the app it came from, so reports about this service sit alongside reports about software that has nothing to do with it. The picture is stored as a file in the same project. Your browser signs in to that project anonymously in order to write the report, and that anonymous sign-in may create its own report and read nothing at all, so no other reporter and no other app can read yours.
Two things read them. The operator, in a private viewer and a command-line tool, both of which require the operator's own Google account. And an automated assistant the operator runs on their own machine to work through the queue, which opens your note, the picture and the technical facts in order to make the fix and to write the reply you get back. That assistant is Anthropic's Claude, so a report being worked on is sent to Anthropic under Anthropic's own terms. It is listed with the other processors below.
Nothing deletes reports on a schedule. A report and its picture stay in that project until the operator removes them, so ask at contact@counteradmin.net and yours is deleted.
Scanning a Menu without an account
You can photograph a Menu and see it read back before you create anything. When you do, the photographs go to the same private storage as an account holder's upload, and the Menu read from them is stored so the result can be shown back to you. Neither is linked to a person.
Two things identify the attempt. A random value in a cookie keeps the result yours when you come back to the page. And to keep one client from spending the service's model budget on its own, your address and that cookie are combined with a server-held secret and stored only as the result of that calculation, which cannot be turned back into an address. The address itself is never written down.
If you create an account, the scan is copied onto your Restaurant as a normal import, which you then review like any other. If you do not, it is deleted automatically 7 days after you made it, and a claimed one 30 days after that. A sweep runs every hour and deletes the stored files before the record, so neither is left behind. That deletion is not conditional on you asking for it.
How long data is kept
| Data | Kept for |
|---|---|
| Account and Organization records | While the account is open, then deleted on request or on closure. |
| Menu content and published Menus | While the account is open. Downgrading a plan never deletes Menu content. |
| Uploaded Menu documents and imports | 30 days after the last time the import was worked on, then the document is deleted automatically. The import itself is kept, so the record that it happened stays with the Menu. There is no button that removes a past import sooner, so ask and it is deleted. |
| Model run metadata | With the import. It contains no Menu text. |
| Scans made without an account | 7 days, or 30 days more once claimed onto a Restaurant. Deleted by an hourly sweep. |
| Reports sent from the button | In the feedback project, with the screen picture, until the operator deletes them. No automatic expiry. |
| Daily counts and Menu visit totals | Indefinitely. They are dates and numbers with nothing to identify. |
| Billing records | As long as tax and accounting law requires, which is normally six years after the final payment. |
| Error reports | The monitoring provider's retention period, at most 90 days. |
Legal bases
- Contract. Running your account, storing your Menu, publishing it, and taking payment.
- Legitimate interests. Keeping the service secure, monitoring errors, preventing abuse of the upload and model endpoints, answering a problem you report, and improving the product. These are balanced against your interests, and none of them involve profiling you.
- Consent. Workspace analytics, and the screen picture in a problem report you choose to send. You can withdraw either at any time.
- Legal obligation. Keeping billing records.
Who else processes your data
These are the only companies that process data on this service's behalf.
| Processor | What it does | Where |
|---|---|---|
| Supabase | Authentication, the application database, and file storage. | European Union |
| Cloudflare | Application hosting and content delivery. It also works out the approximate country and city on a problem report sent by someone who is not signed in. | Global edge network |
| OpenAI | Reading an uploaded Menu document, translating Menu text, and ranking Item photo options. | United States |
| Stripe | Subscription payments. Stripe, not this service, handles card details. | United States and European Union |
| Resend | Sending service email, such as a password reset or an invitation. It sees the address and the message, never your Menu content. | United States and European Union |
| Google (Firebase) | Two separate things: workspace analytics, which stay off until you allow them, and the problem reports you choose to send from the report button. Those reports go to a separate project the operator also collects feedback from other apps in. | United States |
| Anthropic | Triaging a problem report you chose to send, when the operator's automated assistant works the queue. It reads that report, including the screen picture, and nothing else from the service. | United States |
| Sentry | Error monitoring. Reports carry the error and the route, never your content. | United States |
A processor is added here before it is used, not after, and the date at the top changes when one does.
Sending data outside the UK
The table above says where each processor is. Several are in the United States, so data reaches a country whose data protection law the United Kingdom has not found adequate on its own.
Those transfers rely on the standard contractual clauses together with the UK International Data Transfer Addendum, which are the contractual safeguards UK GDPR provides for, and on the transfer terms each of those providers publishes. You can ask for a copy of the safeguards that apply to a particular provider by writing to contact@counteradmin.net.
Security
Every private request is checked against your account, your Organization membership, your Restaurant and your role before it reads or writes anything, so one Restaurant's data is never reachable from another. Traffic uses HTTPS. Private uploads live in a separate storage bucket from published files. Model requests carry no credentials, no storage keys and no identifiers, they are sent without any tool the model could call, and no model response can create, publish or activate anything without a person confirming it first.
If a breach of this service happens that puts your data at risk, you are told about it, and the Information Commissioner's Office (ICO) in the United Kingdom is notified where the law requires it.
Children
This is a product for businesses. You have to be 18 or over to hold an account, and the service is not directed at children and does not knowingly collect anything about one. A published Menu is a public page that anybody can read, and reading it collects nothing that identifies the reader.
If you believe a child's personal data has reached the service, write to contact@counteradmin.net and it is deleted.
Automated decisions
Nothing here makes a decision about you automatically, and nothing profiles you. The machine-assisted features read a Menu document, translate Menu text or rank photo options, and each one produces a draft a person has to confirm before it goes anywhere. None of them looks at you, and none of their output decides anything about your account.
Your rights
Under UK GDPR you have all of these, and none of them costs anything to use:
- Access. Ask what is held about you and get a copy of it.
- Rectification. Have something wrong about you corrected.
- Erasure. Have your data deleted, where nothing else requires it to be kept. Billing records are the usual exception, and the table above says for how long.
- Restriction. Have processing paused while a dispute about it is sorted out.
- Objection. Object to processing that rests on legitimate interests. There is no direct marketing here to object to.
- Portability. Receive the data you gave us in a machine-readable form, or have it sent to somebody else.
- Withdrawing consent. Turn off anything you switched on. Doing so does not undo what was lawful before you withdrew it.
There is no button in the product that exports or deletes an account yet, so these are handled by hand. Write to contact@counteradmin.net and expect an answer within one month. You can correct your own account details and Menu content in the product at any time, and withdraw analytics consent in Settings without asking anyone.
If you are not satisfied with the answer, you can complain to the Information Commissioner's Office (ICO) in the United Kingdom.
Changes
When this policy changes, the version and the date at the top change with it, and sign-up records which version you accepted. A change that affects how your data is used is announced in the workspace before it takes effect.
See also the terms of service.