Legal
Privacy Policy
In force from October 26, 2026. Version 2026-10-26. Your account records the version you accepted.
Last revised October 20, 2026.
What we hold about you, why, who else receives it, how long we keep it, and how to have it corrected or deleted.
Published October 12, 2026. It takes effect on October 26, 2026, which is 14 days later. This version adds the Home Market Packs: the packs you hold and their billing, the checkout of a pack opened and a pack bought, and the offer of a pack shown beside a company outside the United States (clause 3, clause 12). The version before it described the three plans, and what we hold about you includes the plan you pay for beside the plan you use, the answers you gave the guided setup, and what you asked an assistant before you came, when you told us that is how you heard of us. The product still runs in three places on one account, the web and our apps for iOS and Android, and deleting your account is a control in Settings on every one of them. clause 15 describes the one partly automated decision we make about you, which can only ever decide in your favour; every other outcome goes to a person.
Also new here, and each named where it happens: Margin Notes and what is recorded about it (clause 13), the events our own server sends to Google Analytics and which of them are counted when you decline (clause 12), the alerts and the review reminders (clause 13), and the Global waitlist, closed since this version was published, with what it recorded kept (clause 3).
Revised October 20, 2026, after it was published. Three things were added, and none applies to you unless you use it: a row on the browser extension, what it sends to our API and the daily marks of use we keep; a subclause on the Claude connector, what it records if you connect it, what Anthropic receives, how long we keep it and how to end it; and a subclause on the ChatGPT app, which works the same way with OpenAI as the recipient (clause 3). Neither connector can be connected until every account has been emailed about it. And for Margin Notes: you can now switch it on or off in Settings or on a card in the app, and we record each of those choices; its unsubscribe link no longer expires; and one click on it stops all our marketing email (clause 13).
The previous version, in force from September 27, 2026, is replaced by this one in full. The 14 days of notice promised below are given.
If you want only the essentials, read clause 2 and clause 10, and clause 12 for the switch that turns the measurement off.
1Who holds your data, and how to reach us
ValueScreener is operated by Oliva Real Estate Technologies LLC, a limited liability company formed in Delaware, United States. That company is the controller of the personal data described in this policy. Its registered agent for service of process is A Registered Agent, Inc., 8 The Green, Ste A, Dover, DE 19901, United States.
Write to legal@valuescreener.ai for anything in this policy: a copy of your data, a correction, deletion, an objection, or a question. A person reads that mailbox. For everyday product problems, use hello@valuescreener.ai.
2The short version
We sell a research subscription for money. That is the whole business model, so we have no reason to collect more about you than the product needs. We hold your account details, what you write in the product or the support chat, a record of milestones such as a thesis created, what we billed and what we emailed you, and server logs. We do not sell personal data and we run no advertising.
The site can load Google Analytics, Microsoft Clarity and HubSpot, so that we can tell which pages get read, which links bring people here and which parts of the product get used. None of the three sets a cookie, records a session or reads anything in your browser until you say yes; refusing takes one click, is offered as plainly as accepting, and is remembered. One count does not wait for your answer: our own servers send fourteen events about a subscription and its money to Google Analytics even if you decline, with no cookie, no code for you and no session, so that nothing in them can be joined to you or to any other event. Clause 12 names all fourteen. Google Analytics and Clarity also run inside the product once you say yes, under a one-way code made from your account that is not your email or account number. They see which sections and features you use, never what you write, your holdings or any amount, and Clarity masks those on screen. HubSpot's tag stays on the public pages. The Android app asks the same question once, and runs Google Analytics and Clarity only if you say yes there. Declining changes nothing about what the site or the app does for you. clause 12 says what each one collects and how to take your answer back at any time.
HubSpot is also where we keep a record of each account, so that we can see where accounts come from and follow up a trial or a cancellation. It receives that record from our servers, never what you write in the product or which companies you hold. Clause 6 lists exactly what it holds.
If something below is wrong about your account, tell us and we will fix it.
3What we hold
This is the personal data in our own database. Nothing is collected silently that is not listed here.
| Category | Fields | Where it comes from |
|---|---|---|
| Account | Email address and a normalised copy of it, display name, a one-way hash of your password, whether the address is verified, account created and last sign-in times, your time zone if the browser reports a valid one, your light or dark preference, and your preferred markets. Your investing profile, if you answer the guided setup or set it in Settings: experience level, investing style, investing objective, reading mode, preferred currency and markets. | You, at registration and in settings. The investing profile only if you answer it; change or clear it in Settings. |
| Acceptance of these documents | The moment you accepted the terms and this policy, the version string you accepted, and the IP address the acceptance came from. | Recorded automatically when you register. |
| How you found us | Your answer to "how did you hear about us", a referral channel, and a referral code if you arrived through one. Whether you signed up with an email address or with Google; the campaign tags, if any, of the link that first brought you; and a partner’s referral code if you arrived through a partner’s link. If you allowed measurement, the identifiers Google Analytics and HubSpot keep in your browser, read when you register so that your earlier visits can be joined to the account in our CRM. | You, at registration. The answer is optional; the rest comes from the link you followed and, for the identifiers, only with your yes. |
| If an AI assistant sent you | If you answer "AI assistant" to "how did you hear about us" when you register, and choose to tell us: which assistant it was, from a short list or typed in if you pick Other, and what you asked it, up to 280 characters. Both questions are optional, both appear only under that one answer, and neither answer is sent to our analytics tools or to our CRM. | You, at registration, and only if you answer. Leaving both empty changes nothing. |
| Your research | Theses per company: bull case, bear case, notes on the moat, free notes, conviction, status, entry and target prices, entry date, and the guardrails saved on them as an indicator, a comparison and a threshold, with whether each line was our draft, your edit of it or your own. Earlier versions of a thesis are kept when you change it. Watchlists and their items. Saved screens and saved valuation versions. Research notes you write on a company. Portfolios you choose to create, with ticker, exchange, number of shares, cost basis, purchase date and notes, and the transactions you record in them: the date, whether it was a purchase, a sale, a dividend or a split, the shares, the price, the amount, any fees and your note. A practice portfolio, if you build one from your shortlist, is kept the same way and marked as practice. | You, in the product. Positions and cost bases are optional; the product works without them. |
| Alerts | Guardrail breach and recovery records with the value at the time, your in-app alerts and whether you read them, your per-type email preferences, and a delivery log per notification. | Generated by the nightly checks on the guardrails you set. |
| Billing | Your subscription state and trial dates, the customer and subscription identifiers issued by our payment processor, and a record per payment: amount, currency, status and date. The plan you pay for and each change of plan with the day it takes effect. The Home Market Packs you hold, each with its market code, the identifier of its subscription, its status, whether it renews, and the end of the period paid for. For each subscription and each change of plan, the record of your agreement to its renewal: the moment, the plan, the price and first-charge date the page showed, the wording’s version and the IP address it came from. If you ask for the refund of a first payment: when you asked, whether it was granted, and the amount returned; and, on a payment refunded that way, the payment processor’s fingerprint of the card (a code that is the same for the same card and is not the card number), so that a request for a refund on a card already refunded on another account goes to a person. | You and the payment processor. Card numbers never reach us. |
| Trial record | A normalised form of your email address, the payment processor’s customer id, and the date. | Written when a trial starts. Kept after the account is deleted, so that each person gets one trial. |
| Spreadsheet add-in key | A one-way hash of the key, its first few characters so you can recognise it in a workbook, and when it was made. The key itself is shown to you once and is never stored. | You, when you create a key in Settings. |
| Browser extension | The browser extension keeps your key in that browser’s own storage. When you ask it about a company, a fund or an index, it sends our API at valuescreener.ai your own key and the ticker, SEC number (CIK) or index name of what you asked about, with a label saying the request comes from the extension, and nothing else: never the address of the page you are on, anything on it, or your cookies. On our side, at most once a day per account, the product use log records that the key was used and that the extension used it, without the page or the company. Each request also appears in the server logs, like any other. | You, when you press its button or type a ticker into it. |
| Invitations | Your invite code and who invited you, if anyone. For each person you invite: the credit it earned, its state, the payment processor’s entries for it, and the result of our checks against abuse, which compare the two accounts’ email addresses, registration network addresses and cards. | The invitation programme and the payment processor. |
| Email delivery log | One row per message we attempted: the address, the type, the subject, the provider, its message id, whether it was sent or failed, and any error. Where our email provider reports it, when the message was delivered and when it was opened. For onboarding and win-back messages, when a link in them was first followed, which we learn because the link passes through our own address first. | Written by our own mailer, so we can prove what we did and did not send. |
| Product use | A per-account record of milestones, each with its time: a trial started; a thesis or guardrail created, with the company, the indicators, and the conviction and status you set, but never your words; that you added a company to your research list; that a list you checked in the free portfolio audit was carried into your new account, with how many companies but never which; a breach email sent, with the guardrail and the value it reported; a payment, with its amount, plan and any promotion code; a cancellation, with the reason you picked but never your comment; a subscription ended; and, at most once a day, that the disciplines panel was shown on a company page or on Today, without the company. Since the guided setup: that you finished or skipped it, and each answer you gave in it; that you reached a first insight, a first shortlist and a first weekly routine, with the days since you signed up; a change of reading mode; a shortlist saved or a watchlist imported, with how many companies and the kind of file but never which companies; a change of plan; a refund; joining the Global list, with how many markets; the checkout of a Home Market Pack opened and a pack bought, with the pack’s market code. And when you last opened your watchlist, your saved screens, the research desk and the weekly brief (one entry per place, routine:<place>), and, during a trial, each day you read the research desk (routine:<date>:research): that is how the weekly-routine milestone and the plan that fits are worked out. Used to tell an active account from a dormant one and to decide what to build. | Generated as you use the product, on the web or in either of our apps, and kept in our own database. |
| Support chat | What you write in the chat and the replies you get; your name and email address, which the chat asks for before your first message so that a reply can reach you; the page you wrote from; and a one-way code made from your network address, used only to stop abuse. Linked to your account if you write while signed in. | You, in the chat panel on any page. |
| Reports of a wrong figure | The company and the figure you reported, what the page showed and the source it cited, the page you were on, what you wrote, your email address if you gave one, your account if you were signed in, and a code made one way from your network address, used only to limit how many reports one address can send. Then our handling of it: its status, its severity and our notes. The correction we publish carries none of this. | You, through the report link on a figure or a page. |
| If you cancel | The reason you select and any comment you write, with the date. If you answer neither, the row records that the question went unanswered. | You, in the cancellation flow. Both are optional. |
| Answers to a one-question poll | The choice you pick and any comment you write, which page the question was on, the date, and your account and plan if you were signed in. No network address is kept with an answer. | You, on the pricing page or after a payment. Answering is optional. |
| Server logs | Request paths, methods, timestamps, status codes, response times and IP addresses. | Generated by the server, for security and debugging. |
| Newsletter list | For someone who joins from a form on this site: the email address, which page it was given on, the campaign link if there was one, the network address it came from, and the date. No name, and no country: we do not work out where you are from your address. Older entries, from the earlier product, may also carry a first name and a country. Margin Notes is sent to it. The address, the page and the campaign tags are also kept in our CRM, HubSpot, as a newsletter subscriber. | You, from the sign-up form on our public pages, and, for the older entries, people who signed up on the earlier product’s pages. |
| Margin Notes records | For each address that receives Margin Notes, or was offered it at sign-up: how it came to (the form, the sign-up notice, the card in the app or Settings), the version of the wording shown, the time of the consent or refusal and, when it was given on a form or at sign-up, the network address it came from, each issue sent to it with its delivery status, and which issues were opened and which of their links were followed, as our email provider records them. After an unsubscribe or a spam report, a one-way fingerprint of the address. | You, when you sign up, answer the card in the app, switch it in Settings or unsubscribe; and our sending system. |
| Global waitlist | Your email address, or your account if you are signed in; the markets you named; whether you would pay monthly or yearly; a note of up to 500 characters if you write one; the page you joined from; the date. The address and the markets are also kept in our CRM, HubSpot, as a waitlist entry and not as a newsletter subscriber. | You, on the waitlist form, while it was open. The Global plan is retired and the list is closed: nothing new is recorded, and what was recorded is kept as this row says. |
| Limits on public forms | One entry per request to a form that takes no account, such as the Global waitlist while it was open: the time and, for a visitor without an account, a keyed hash of the network address (a code made with a secret key of ours, which cannot be turned back into the address without that key), or the account for one who is signed in. Never the address itself. | Written by the server when the form is used, to stop one sender filling it many times. |
| People named in company research | For a company we cover, the name and title of each executive its proxy statement names, as the filing prints them, and the pay the company reported for each, with the filing every figure was read from. On the analyst panel, the names of the firms and analysts behind a forecast. None of it is about you unless you are one of them. | Proxy statements filed with the SEC, as our market data supplier reads them, and that supplier’s analyst records. Nothing in this row comes from you. |
Our apps for iOS and for Android use the same account and the same data as the web. On the phone each keeps your sign-in, and an offline copy of your theses, inbox and tracker, which the phone never includes in its backups. Neither app reads your contacts or your location, and neither uses an advertising identifier or a device fingerprint. Measurement in the apps waits for the same yes as on the site, and clause 12 says what the Android app sends if you allow it. The app stores distribute the apps and keep their own records of your download under their own policies, not this one.
If someone invited you, we tell them when your first payment goes through, when their credit lands and, if it is later reversed, whether a refund or a disputed payment was the reason. Never your name or your email address.
3.1What we deliberately do not collect
- No brokerage or bank credentials, and no connection to any brokerage account. There is no field for them and no integration that could ask.
- No card numbers. Card details are entered on our payment processor and never touch our servers.
- No copy of a file you import. A broker export or a list you upload is read in memory to find the companies and positions in it, and the file itself is not kept.
- No government identifiers, no date of birth and no phone number.
- No special category data: nothing about health, beliefs, politics, ethnicity or sexual life. Do not put such data into a free text field.
- No account for the free public pages. The company catalogue is served by a public endpoint that returns company figures only and holds nothing about you, and the calculators run in your browser, so what you type in them never reaches us.
- No advertising or fingerprinting scripts anywhere. The measurement tools named in clause 12 run only after you agree, on the site and in the Android app, and none of them is connected to any advertising feature or account.
3.2The Claude connector, if you connect it
The connector lets Claude, the assistant made by Anthropic, read company figures from ValueScreener and, if you allow it, your own watchlist and theses. Nothing in this subclause exists for you until you connect it, from Claude by signing in here, or from Claude Code with your personal key.
| Category | Fields | Where it comes from |
|---|---|---|
| Claude connections | For each connection you allow: the client that asked (the name and return address it registered), what you allowed it to read, when you allowed it, and its access and refresh tokens, stored only as one-way hashes with their first few characters, with when each expires, when it was last used and when it was ended. A sign-in waiting for your answer, and a one-time code that lives for minutes, are kept the same way. | You, on the consent screen; and Claude, when it registers and signs in. |
| Connector use | One line a day in the product use log saying that your account used the connector. In the server logs, the name of each tool Claude called and how long it took. Never what it asked: no question, company or other argument is recorded. | Written by our server when Claude calls it. |
- Why: to answer the calls you asked Claude to make, to make sure a connection is used only by the account that allowed it, to keep each connection within its limits, and to list your connections in Settings so you can end them. The legal basis is our contract with you: you asked for the connection.
- Who receives it: Anthropic receives the answers our server gives to Claude on your connection. That is company figures and the filings behind them and, if you allowed it, your watchlist with your notes on it, and your theses with their status, entry and target prices and guardrails. It receives them because you asked Claude for them, and holds them under its own terms and privacy policy, not this one: Anthropic is not our processor, and we have no contract with it about your data. We receive only the call Claude makes (a tool and what it names, such as a ticker), never the conversation around it, and we send Anthropic nothing about you beyond those answers.
- How long: a token is deleted 30 days after it expires or the connection is ended; a sign-in waiting for an answer and a one-time code are deleted a day after they expire; a client registration that no connection uses any more is deleted 30 days after it was made. A job deletes them every night. Deleting your account deletes its connections at once.
- Your controls: end any connection in Settings, Claude and ChatGPT, and its tokens stop working on the next call. Replace or revoke your personal key in Settings to cut off Claude Code. You can also remove the connector in Claude, which stops Claude calling us but leaves the records above to be deleted on the schedule above.
3.3The ChatGPT app, if you connect it
The ChatGPT app is the same connector, answering ChatGPT, the assistant made by OpenAI. Nothing in this subclause exists for you until you connect it from ChatGPT by signing in here. What a connection records, why, how long we keep it and how to end it are as in the subclause above, with ChatGPT in place of Claude. One thing differs:
- Who receives it: OpenAI receives the answers our server gives to ChatGPT on your connection. That is company figures and the filings behind them and, if you allowed it, your watchlist with your notes on it, and your theses with their status, entry and target prices and guardrails. It receives them because you asked ChatGPT for them, and holds them under its own terms and privacy policy, not this one: OpenAI is not our processor, and we have no contract with it about your data. We receive only the call ChatGPT makes (a tool and what it names, such as a ticker), never the conversation around it, and we send OpenAI nothing about you beyond those answers.
- Your controls: end the connection in Settings, Claude and ChatGPT, or remove the app in ChatGPT, which stops ChatGPT calling us but leaves the records to be deleted on the schedule above.
4Why we hold it, and on what legal basis
For readers in the European Union, the European Economic Area and the United Kingdom, the legal bases are those in Article 6 of the GDPR.
| What we do with it | Data used | Legal basis |
|---|---|---|
| Run your account: authenticate you, keep you signed in, show you your own work. | Account, research. | Performance of our contract with you. |
| Do the thing the product is for: re-test your guardrails against new data on a schedule, and send you the alerts, filing and saved-screen notices, inbox items and weekly brief that follow. | Research, alerts, account, email preferences. | Performance of our contract with you. |
| Take payment, prevent payment fraud, issue invoices and keep accounts. | Billing, account. | Performance of our contract, and compliance with tax and accounting law. |
| Send transactional email: verification, password changes, receipts, a failed payment, the trial ending or ended, a cancellation, a deleted account. | Account, billing, email log. | Performance of our contract. |
| Send product email about the tool itself, and the onboarding sequence that follows registration. | Email address, display name. | Our legitimate interest in helping a new subscriber use what they signed up for. One click in any of these messages stops them. |
| Tell our own team when someone registers, adds a first company, pays or cancels. | Email address, display name, plan and, for a payment, its amount. | Our legitimate interest in running the service. |
| Give each person one trial. | Trial record. | Our legitimate interest in preventing abuse of the trial. |
| Run the invitation programme: credit the person who invited you, tell them what happened to the credit, and stop abuse of it. | Invitations, billing, and the network address recorded when each account accepted these documents. | Our contract with the person who invited, and our legitimate interest in preventing abuse. |
| Learn which AI assistants send readers to us and what they were looking for, so that we can describe the product accurately where people ask about it. | If an AI assistant sent you. | Our legitimate interest in knowing where our readers come from. Both answers are optional, nothing depends on them, and neither leaves our own database: no analytics tool and no CRM receives either. |
| Send Margin Notes to people who signed up for it on one of our forms, or switched it on with the card in the app or in Settings. | Newsletter list; email address. | Your consent, given on the form, on the card or in Settings. Withdraw it at any time with one click in any issue or, with an account, in Settings. |
| Send Margin Notes to accounts opened after the sign-up form began offering it, that did not refuse it there. | Email address. | Our legitimate interest in writing to our own customers about the service they use, which the law allows for existing customers (Article 6(1)(f) GDPR; article 21.2 of Spain’s Law 34/2002). You can object at any time, free, with one click in any issue or in Settings. |
| Answer what you write when you reply to Margin Notes. | Your address and what you write. | Our legitimate interest in answering you. |
| Know whether Margin Notes and our other emails arrive and are opened, and which of their links are followed, to fix what does not arrive and to learn what readers find useful. | Your address; which messages were opened and which links were followed, with the time, as our email provider records them. | Our legitimate interest in knowing that our email arrives and is read, and in making it more useful. It is never used to build a profile of you. You can object to it at any time, free: one click on the unsubscribe link in any issue of Margin Notes, or its switch in Settings, stops the letter and the record of its opens and clicks with it. For our other emails, the ways to stop the record are in the email clause, and you can write to legal@valuescreener.ai. |
| Write to a creator, publisher, association or society, at their work address, to ask whether they want to be a partner; and keep enough of a record to know what we sent, what they answered, and never to write to them twice. | Prospect record: work address, name where we have it, organisation, role, public web address, the sentence about their own published work, the messages sent and any reply. | Our legitimate interest in finding partners for a paid research tool, balanced against the recipient of a message they did not ask for. We write to a work address about their professional role, never to a private one about their private life; we say in the message where we found them; one reply saying no, or one click, ends it for good. Object at any time, free: reply to any message, or write to the address in the outreach clause. In the United Kingdom, Germany and anywhere else whose law requires consent before a message of this kind, see that clause. |
| Stop all email to an address that reported us as spam, from any of our addresses. | A one-way fingerprint of the address. | Our legitimate interest in respecting that request and in keeping our email deliverable. |
| Avoid sending an old newsletter sign-up a second welcome sequence when they register. | Newsletter list. | Our legitimate interest in not emailing you twice. |
| Keep the service secure, investigate abuse, enforce rate limits, and hold evidence of what was agreed. | Server logs, limits on public forms, acceptance record, account. | Our legitimate interest in a secure service and in being able to prove the terms of a contract. |
| Understand whether accounts are being used, and decide what to build or fix. | Product use events, aggregate counts. | Our legitimate interest in improving a product we sell. Never combined with data from anyone else and never shared for advertising. |
| Improve the product from what people tell us when they leave. | Cancellation reason and comment. | Our legitimate interest in a better product. Both are optional, and answering is never a condition of cancelling. |
| Find out what holds people back from subscribing, and what almost did, so we can fix it. | Answers to a one-question poll. | Our legitimate interest in a better product. Answering is optional, and nothing depends on it. |
| Answer what you ask in the support chat, and reach you with the reply if you leave before it arrives. | Support chat. | Our legitimate interest in answering you, and our contract with you if you subscribe. |
| Recognise the key a workbook or the browser extension sends and answer it with company figures. | Spreadsheet add-in key; browser extension. | Performance of our contract with you. |
| Read a report that a figure is wrong, write back to the person who sent it if they left an address, and publish the correction, which carries nothing about that person. | Reports of a wrong figure. | Our legitimate interest in correcting what we publish, and in answering the person who told us. |
| Publish who runs a company we cover and what it reported paying them, and who published a forecast about it. | People named in company research. | Our legitimate interest in publishing research on companies from what they file and what is published about them. A person named can use the rights in clause 10. |
| Choose, from the answers you gave in the guided setup (objective, experience, style and reading mode), where you land after signing in, which ready-made screen opens first and how much each page explains. Everybody who opens the same company sees the same facts; only the explanation around them and the starting point change. | Investing profile. | Performance of our contract with you. Skipping a question, or changing or clearing an answer in Settings, leaves the standard starting point and reading. The answers are kept with the account and deleted with it. |
| In the last three days of a trial, show you which plan covers what you used in it, marked "Fits what you used", with your own counts beside it. | Read from the product use log and what the account holds, counted inside the trial: the days you read the research desk, the days you opened the tracker, the days the spreadsheet key was used, weekly briefs opened, guardrails on your theses, screens watched overnight, holdings, saved valuations and research notes. Counts only, never the companies. | Our legitimate interest in telling you, before the trial ends, what each plan would keep. It decides nothing: Curious, the free plan, is the answer when it fits, and the choice is always yours. |
| Write to you once if we build coverage for the markets you named, and count how many ask. Never marketing. | Global waitlist. | Your consent, given by submitting the form. Withdraw it at any time by writing to us, and we delete the entry. |
| Keep one record of each account in our CRM, HubSpot: where it came from, how its trial and subscription are going, and why it ended if it did, so we can follow up a trial, read a cancellation, and tell which campaigns bring readers who stay. | Your email address and name, account and billing dates, plan and monthly amount, failed payments, cancellation reason, how many theses and holdings you keep, your investing profile, the dates you first reached the three first-use milestones, and the campaign tags of your first and latest visits, as clause 6 lists them. Never what you write in the product, and never which companies you hold. | Our legitimate interest in running and selling the service. You can object; see clause 10. |
| See which pages get read, which links bring people here, where the signup flow loses people, and which parts of the product are used and which are not, so we can spend a small budget on what works and build what readers actually use. | The measurement data listed in clause 12: pages viewed, the actions listed there, approximate location from the IP address, device and browser, how you reached the site, and which campaign first brought you. Google and Microsoft never join it to your account. If you open an account, the pages HubSpot's tag recorded are joined to your record in our CRM, and the record carries your Google Analytics identifier. Inside the product, Google and Microsoft receive a one-way code made from your account (never your email, name or account number) so that your visits on two devices count as one reader; they cannot turn it back into your account, and we never send them anything that could. | Your consent, and nothing else. We do not claim legitimate interest for this: it needs consent to store anything in your browser, and using a different basis for the processing that follows would be a distinction without a difference. Withdraw it in clause 12 at any time. |
You can object to any processing we base on legitimate interest. See clause 10.
5What we will never do with it
We do not sell personal data, we do not share it for cross-context behavioural advertising, and we do not disclose it to data brokers.
- We do not use your writing to train machine learning models, and we do not send it to anyone who would.
- We do not show your theses, notes or positions to other users, and there is no public profile.
- We do not build profiles about you for anyone else, and we do not enrich your record from third-party data sets.
- We do not use your data to make decisions about you beyond running the product you paid for.
Two ordinary exceptions, stated plainly. We disclose data where the law compels it, and we will tell you unless we are forbidden to. And if the business is sold or reorganised, the data moves with it, bound by this policy until it is replaced by a policy no less protective; we will tell you before that happens.
6Who processes it for us
These are the third parties that receive personal data from us, and what each receives. Each acts under a contract with us, except Telegram, with which we have no data processing agreement.
| Processor | What it does | What it receives | Where |
|---|---|---|---|
| Stripe, Inc. | Payments, checkout, invoices, the billing portal, tax calculation. | Your email address, name, billing address and card details, which you give it directly, plus the subscription record. | United States, with global processing. |
| Resend | Delivery of transactional email and of Margin Notes, and the record of whether each message was opened and which of its links were followed. | Your email address, name, the content of the message, and when it was opened and which links in it were followed. | United States. |
| Brevo | Delivery of product and onboarding email. | Your email address and name. | European Union. |
| Cloudflare, Inc. | DNS, content delivery, denial-of-service protection, and the encrypted tunnel through which the site is reached, and the routing of email sent to our addresses, including replies to Margin Notes, to the mailbox where they are read. | Request metadata in transit, including your IP address; and the email you send to our addresses. | Global edge network. |
| Google LLC (Gmail) | The mailbox where replies to Margin Notes, and other email to our addresses, are read. | Your email address and what you write. | United States. |
| Slack Technologies | Internal notices, so a small team sees registrations, a first company added, payments, failed payments and cancellations. | Your email address, display name and plan, and for a payment its amount. | United States. |
| Telegram | Carries support chat messages to the person who answers, and the replies back. Also tells us when a subscription starts or ends. | What you write in the support chat, with your name, email address and the page you wrote from. When a subscription starts or ends, your email address, and for a start the amount and plan. | Outside the EEA, with no transfer agreement: see clause 7. |
| Google Ireland Limited (Google Analytics 4 and Google Tag Manager) | Counts visits to the site and use of the product, and tells us which link or search brought them. Tag Manager is the loader the tag arrives through; it collects nothing of its own. | Only if you agree first: pages viewed on the public site and in the product, the actions listed in clause 12, the referring link, a randomly generated identifier stored in your browser, when you are signed in a one-way code made from your account together with your plan, trial state and how long ago you joined, and, if you answered the guided setup, your experience level, investing style, objective and reading mode as broad labels, the plan you pay for, your approximate location derived from your IP address, and your device and browser. Never your name, your email, your account number, or anything you write in the product. | European Union and United States. |
| Microsoft Ireland Operations Limited (Microsoft Clarity) | Aggregate heatmaps and session replays of the public pages and the product, so we can see where a page or a feature confuses people. | Only if you agree first: a recording of how you use the site, meaning clicks, scrolling, mouse movement and what the page shows, plus the same page, device and approximate-location information and, when you are signed in, the same one-way code and plan Google receives and, if you answered the setup, your experience level and reading mode as broad labels. Text typed into a field is masked by the tool, and in the product we also mask your holdings, amounts, transactions, theses, weekly brief, name and email address, so they never appear in a recording. It never runs on the signup, sign-in or payment screens. | European Union and United States. |
| HubSpot, Inc. (HubSpot CRM, and its tracking code) | Our customer relationship system: one record per account, so we can see where accounts come from and follow up a trial, a failed payment or a cancellation. Its tracking code counts which public pages get read. | About account holders, from our servers: your email address, your name if we hold it, when the account was made and when you signed up, the state and dates of your trial and subscription, your plan and what it brings in a month, the plan you pay for and the plan you can use today, your answers to the guided setup (experience level, investing style, objective and reading mode), the dates you reached the three first-use milestones, whether you joined the Global list while it was open and for which markets, how many payments failed, your cancellation reason and date, and how many theses and holdings you keep, never what they say or which companies they are. Also the campaign tags of your first and latest visits and, if you allowed measurement, your Google Analytics identifier and HubSpot's own browser identifier. About newsletter subscribers, from our servers: the email address, the page it was given on and the campaign tags of the link, if any. About someone who joined the Global list without an account while it was open: the email address, the markets named and the date, as a waitlist entry, never as a newsletter subscriber. About a partner whose link brought you: their code, on your record, and on theirs how many accounts it brought and how many pay, never who. About a person we wrote to about a partnership who never asked us to, which is a different thing and has its own clause 14: their work address, their organisation and role, the campaign, where the sequence has reached, what any reply was taken to mean, and whether they told us to stop. Only if you agree first, from your browser: the pages listed in clause 12, joined to that record if you open an account. Never your password, what you write in the product, your holdings, a ticker or your card. | European Union, in HubSpot's EU data centre, with access from the United States: see clause 7. |
| Smartlead.ai (Infynd Pte. Ltd.) | The tool that sends the partnership messages described in the outreach clause, and holds the replies to them. It is never used for any email to a reader, a subscriber or a customer. | About a person we wrote to about a partnership who never asked us to: their work address, their name where we have it, their organisation and role, its public web address, the one sentence we wrote about their own published work, which of our three messages were sent and when, and anything they wrote back. Nothing about any reader, subscriber or customer of ours, and nothing from the product. See clause 14. | United States. |
Google, Microsoft and HubSpot's tracking code are different from the others in one way that matters: they are the only ones that need your permission, because they are the only ones that put something in your browser for a purpose other than delivering the service. Say no, and none of them is contacted from your browser at all. HubSpot still receives the account record above from our servers, because keeping it puts nothing in your browser. We do not enable the advertising features of any of the three, we have not linked them to any advertising account, and we do not allow any of them to use what it collects here for its own advertising.
The market data sources and public statistical agencies we read receive company identifiers only. They are never told anything about you, and they are not processors of your personal data.
Anthropic, which makes Claude, and OpenAI, which makes ChatGPT, are not on this list. If you connect one of their assistants, its maker receives the answers to the calls you ask that assistant to make, under its own terms and not as our processor; clause 3 says what those answers carry.
7Where your data is held, and transfers out of the EEA
We are unusual here, so it is worth being exact. The application and its PostgreSQL database do not run on rented cloud infrastructure. They run on hardware we own and administer, in our own office in Spain, reachable only through an encrypted tunnel. Your account data and everything you write in the product are therefore processed in the European Union by default.
Some of the processors in clause 6 are established in the United States, so using the service involves transfers there. Those transfers rely on the European Commission standard contractual clauses, with the United Kingdom addendum where it applies, and on the EU-US Data Privacy Framework where the recipient is certified under it. We keep the volume of data that leaves the EEA small by design: what crosses is your email address and name, the emails we send you, the billing record held by the payment processor, the plan and amount in our internal notices, request metadata at the network edge, and, if you agreed to it, the public-page measurement data in clause 12. Google and Microsoft both contract through their Irish entities and both are certified under the Data Privacy Framework, and both may process onward in the United States. If you would rather no data of yours left the EEA for that purpose, decline, and none does.
HubSpot holds the record of your account in its European Union data centre. It is run by HubSpot, Inc., a United States company that is certified under the Data Privacy Framework and whose data processing agreement incorporates the standard contractual clauses, and HubSpot may reach that record from the United States under those safeguards. That is a transfer even though the record stays in the EU, so it is listed here.
Telegram is the exception. It is established outside the EEA, and it is covered by neither standard contractual clauses nor an adequacy decision. What you write in the support chat passes through it, with your name, email address and the page you wrote from, and so does our notice when a subscription starts or ends, with the subscriber's email address. If you would rather your words did not pass through Telegram, write to hello@valuescreener.ai instead of using the chat.
Nightly snapshots of the database are kept on the same hardware.
You can ask us for a copy of the transfer safeguards we rely on. Write to legal@valuescreener.ai.
8How long we keep it
| Data | Kept for |
|---|---|
| Account record and everything you wrote in the product: theses and their earlier versions, guardrails, notes, research notes on companies, watchlists, saved screens, saved valuations, portfolios (a practice portfolio included) and their transactions, and your investing profile. | Until you delete it, or until the account is deleted, whichever comes first. Cancelling does not delete it. |
| Alerts, notification preferences and the product use log. | Deleted with the account. |
| Invitations. | Deleted with the account of either person. |
| Trial record. | Kept after the account is deleted, with no end date, so that the same person cannot take a second trial. |
| Support chat. | Until you ask us to delete it. Deleting an account removes its link to the chat but not the chat itself, so ask, and we will. |
| Newsletter list. | Until you ask us to delete your entry. Every issue of Margin Notes carries a one-click unsubscribe, and you can also write to legal@valuescreener.ai. |
| Margin Notes records. | The record of how an address came to receive the letter (the form or sign-up notice shown, and the consent or the refusal) and of which issues were sent to it: three years after the last issue sent to it, so that we can show we were allowed to send it. After an unsubscribe, a one-way fingerprint of the address, kept so that it is never added back by mistake, including after an account is deleted. The record of which issues were opened and which links were followed: 13 months. |
| Prospect record (a person we wrote to about a partnership who never asked us to). | Two years from the last message we sent them, then deleted, unless they became a partner or a customer, in which case the record of the partnership or the account governs. If they tell us to stop, or report a message as spam, we keep the address and nothing else, with no end date, so that it is never written to again. Ask us to delete it sooner at legal@valuescreener.ai and we will, keeping only that do-not-write entry. |
| Limits on public forms. | About a day: entries more than a day old are deleted as new requests to the form arrive. |
| Global waitlist. | With an account: deleted with the account. Without one: until you ask us to delete it, or until the waitlist ends, which is when we have built the coverage and written to you, or have decided not to build it and said so. The waitlist entry in our CRM goes at the same time. |
| If an AI assistant sent you. | Kept with your account, and deleted when your account is deleted. |
| Cancellation reason and comment. | Deleted with the account. |
| Answers to a one-question poll. | Deleted with the account when one is attached. An answer given while signed out carries no account and no network address, so it is linked to nobody. |
| Spreadsheet add-in key. | Until you replace or revoke it in Settings, or the account is deleted. |
| Claude connector and ChatGPT app connections. | A token 30 days after it expires or the connection is ended; a sign-in waiting for an answer and a one-time code a day after they expire; a client registration no connection uses, 30 days after it was made. Deleted with the account at once. |
| Reports of a wrong figure. | Kept after the account that sent one is deleted, with only the link to the account removed, because the correction a report caused outlives the account and the two have to stay checkable against each other. Ask us and we remove your address and anything that identifies you in what you wrote. |
| People named in company research. | Reported pay for the latest ten fiscal years of each company we cover, replaced as newer proxy statements are read. |
| Billing records: payments, invoices, subscription history. | As long as tax and accounting law requires, normally seven years, even after the account is deleted. We keep the amount, the date and the identifiers, and on a refunded payment the card’s fingerprint, never a card number. A refund claim under the alerting commitment is a billing record too: the claim, what we decided, and the record of what we had sent you that the decision was read from. |
| The record that you accepted these documents, and which version. | While the account exists, and afterwards only where we need it as evidence of a contract that is still capable of being disputed. |
| Email delivery log. | Kept while the account exists, and kept afterwards only where the entry is still needed to show what we sent or did not send. You may ask us to remove your entries. |
| Server logs. | Rotated every night and kept for about five days of generations, then discarded. Never archived. |
| Database snapshots. | Fourteen days, then overwritten. A deletion is therefore fully worked through the snapshot set within fourteen days. |
| Arrivals from a campaign link. | Indefinitely, as a count. One row per arrival carrying a channel name and a time, with no identifier of any kind: no address, no browser string, nothing that could be traced back to a person. |
| The record of your account in our CRM, HubSpot. | While the account exists. Deleting the account deletes the record, and the pages joined to it, there too. Ask us and we will do it sooner. |
| Public-page measurement, if you agreed to it. | Held by Google, by Microsoft and by HubSpot, not by us. We read their reports and copy nothing into our own database. Google Analytics is set to delete user-level and event-level data after 14 months, keeping only aggregate totals after that; Microsoft deletes Clarity recordings and metrics on its own published schedule; HubSpot keeps the pages its tag recorded with the record they are joined to, and otherwise on its own published schedule. Withdrawing your consent stops collection immediately, and if you want what is already there erased, ask us and we will ask them. |
9How we protect it
This list is what we actually do. We claim no certification we do not hold, and no measure we have not implemented.
- Passwords are hashed with bcrypt at twelve rounds and never stored or logged in any recoverable form. We cannot tell you your password because we do not have it.
- Access tokens are short-lived. The longer-lived refresh token is stored on our side only as a SHA-256 hash, is single-use, is rotated on every refresh, and is revoked on sign-out, so a stolen token has a narrow window and cannot be replayed.
- Sign-in is locked out after repeated failures from an address, registration is rate limited, and the API is rate limited per account.
- Traffic is encrypted in transit. The origin server accepts no direct inbound connections from the internet: it is reachable only through an authenticated outbound tunnel.
- Browser requests are accepted only from an explicit allowlist of our own origins, and responses carry hardening headers: no MIME sniffing, no framing, a restrictive referrer policy, no access to camera, microphone or location, and strict transport security in production.
- Text you submit is length-limited and sanitised, request bodies are capped, and queries are parameterised by the data layer, so submitted content cannot be executed as code or SQL.
- The database runs on hardware we administer ourselves rather than a shared managed service, and administrative access to it is limited to the operator of the service.
- Snapshots are taken nightly, and restoring one is rehearsed on a throwaway copy.
No system is perfectly secure. If you find a weakness, write to legal@valuescreener.ai and we will take it seriously, act on it, and thank you.
10Your rights, and how to use them
If you are in the European Union, the European Economic Area or the United Kingdom you have the rights below. We extend the same rights to everyone else, because running two standards would be more work than doing it properly once.
- Access. Ask what we hold and get a copy.
- Portability. Get the data you gave us in a machine-readable form. The product exports screener results and portfolio holdings as CSV by itself; for a full copy of the account, ask us and we will assemble it.
- Rectification. Have anything inaccurate corrected. Your preferences and everything you wrote can be edited in the product, and your name in either of our apps. Your email is changed by us when you ask.
- Erasure. Have the account and its contents deleted, subject to the billing records tax law requires and the trial record. This one you do not have to ask for: it is a control in Settings on the web and in both apps.
- Restriction and objection. Object to processing we base on legitimate interest, including the product use log and the product email sequence, and ask us to restrict processing while a dispute is open.
- Withdraw consent. Where we rely on consent, take it back at any time, with no effect on what was lawful before.
- Complain. Complain to your national data protection authority. In Spain, where the servers sit, that is the Agencia Española de Protección de Datos.
10.1How to exercise them
Write to legal@valuescreener.ai from the address on the account, or from another address if you tell us which account you mean. We may ask a question to satisfy ourselves that the request is really yours; we will not ask for a document you have no reason to send us. We answer within 30 days, and inside one month at the outside as the GDPR requires. There is no charge unless a request is excessive or repetitive, in which case we will say so before doing anything.
Cancelling a subscription is self-serve, in Settings. So is deleting the account, and on all three clients: in Settings on the web, and in Settings in the iOS and Android apps. The screen lists what goes and what we keep before you confirm, and it is the same deletion whichever one you use. You can still write to that mailbox instead and a person will do it. Deleting ends any subscription at the same moment. Once done, the account and your research are gone from the live database at once, except the trial record, and out of the snapshot set within fourteen days.
11California residents
This clause is for residents of California and uses the vocabulary of the California Consumer Privacy Act as amended by the CPRA.
We have not sold personal information, and we have not shared it for cross-context behavioural advertising, in the past twelve months, and we do not do either today. We do not offer financial incentives for personal information.
That statement covers the measurement tools in clause 12, and it is a statement about how they are configured rather than about what they are capable of. The advertising features of Google Analytics are switched off, no advertising account is linked to the property, and the site denies the advertising consent signals in every case and never grants them. Analytics used only to count readers of our own pages is not a sale and is not cross-context behavioural advertising; using the same product with those features on could be, which is why they are off and why this sentence names them.
11.1Categories involved
Identifiers, such as your name, email address and IP address. Commercial information, such as your subscription and payment history. Internet or network activity, such as the product use log and server logs. Financial information limited to the subscription record; the payment card itself is held by our payment processor and not by us. Everything you write in the product, which is content you chose to create. Your investing profile and any Global waitlist entry, which is information you provide. We do not collect sensitive personal information as that term is defined, and we do not use or disclose personal information for any purpose other than those listed in clause 4.
11.2Your rights
You may ask us to tell you the categories and specific pieces of personal information we hold, the sources, our purposes and the categories of recipients; to delete it; and to correct it. Because we neither sell nor share personal information, and because we do not collect sensitive personal information, the rights to opt out of sale or sharing and to limit the use of sensitive personal information have nothing to operate on, but you may exercise them and we will confirm that position in writing.
Use legal@valuescreener.ai. An authorised agent may act for you with written permission we can verify. We will never deny you service, charge you a different price, or give you a lesser product because you exercised a privacy right.
13Email we send, and how to stop it
Four kinds, and the difference matters.
- Transactional. Verify your address, password changes, receipts, a failed payment, the trial ending or ended, a cancellation, a deleted account, and, if you invite someone, what happened to the credit. These are part of the service and cannot be turned off while the account exists.
- The service itself. Guardrail breach alerts; a notice when a company you hold, watch or wrote a thesis on files a delisting notice, sells shares outside a registered offering, changes its auditor, says its past statements can no longer be relied on, files for bankruptcy or receivership, or reports a debt accelerated or increased by a triggering event (every other material filing is listed in Alerts in the product, without an email); a reminder to review a thesis when a quarter has passed, after the company files a periodic report, or when a written reason you set a review date for reaches it, at most one a night; the overnight changes on a saved screen you watch; and the optional weekly brief. This is what you subscribed for. Four of the five have a switch in settings and an unsubscribe link that works without signing in: the filing notices, the review reminder, the saved-screen changes and the weekly brief. The breach alert is the fifth, and it deliberately has neither. The guardrail is its switch: you set it, for this one purpose, and removing it is what stops the alert.
- Margin Notes. At most one email a week, from our founder, about what changed in the filings of US-listed companies. It is the same letter for every reader and says nothing about your own holdings. You receive it if you signed up for it on one of our forms, if you switched it on with the card in the app or in Settings, or if you opened your account after the sign-up form began offering it and did not tick the box to refuse it. Every issue has a one-click unsubscribe that works without signing in, and that one click stops all marketing email to that address, the product email below included; it never stops an alert or other service email. With an account you can instead switch the letter off in Settings, which stops only the letter. You can also reply to an issue, or write to legal@valuescreener.ai. If you report an issue as spam, we stop all email to that address from any of our addresses, service email included, until you write to us to restore it.
- Product email. A short sequence after you register that explains how to use the tool, and occasional notes about the product. One click on the unsubscribe link in any of them stops all marketing email to that address, Margin Notes included.
Unsubscribe links are signed, so nobody can unsubscribe you by guessing a URL. The Margin Notes link does not expire, so an old issue still lets you leave. We log every message we attempt, which is how we can tell you exactly what we sent and when.
Opens and clicks. Margin Notes and our other emails may record, through our email provider, whether a message was opened and which of its links were followed: a small image in the message reports the open when your mail program loads it, and a link passes through the provider on its way to our site. We use this for knowing that our email arrives and is read, for fixing what does not arrive, and for learning which parts readers find useful, never to build a profile of you. To stop it: unsubscribe from Margin Notes with the link in any issue or the switch in Settings, and switch off the other emails that have a switch there; set your mail program not to load images, which stops the open being recorded; and reach our site by typing its address rather than following a link in an email, which stops the click being recorded. The transactional emails cannot be switched off while the account exists.
14If we wrote to you about a partnership and you never asked us to
This clause is not about readers, subscribers or customers. It is about one other group of people, and if you are one of them you probably arrived here from a link at the foot of an email from us. It says what we hold about you, where we got it, why we think we are allowed to, and how to make it stop. Everything in it is true of what our two tools actually hold; nothing here describes an intention.
Who. People who publish about investing to an audience of their own: writers, newsletter authors, podcast hosts, course authors, people who run investor associations, societies, clubs and meetups, and the organisations behind them. We write to ask whether they want to be a partner, which means recommending ValueScreener to their audience for a share of what those readers pay. We do not write to private individuals about their own investing, ever, and we do not sell anything in these messages.
What we hold. A work email address; a name where the address or the organisation's own pages gave us one; the organisation, its public web address and the role we believe the person holds; one sentence we wrote about their own published work, with the public page we read it on; which of our three messages were sent and on what day; whether a message bounced, was replied to, or was unsubscribed from; and, where a reply was read by a person here, what it was taken to mean. In our customer system the same record also carries where in the sequence it has reached and whether the person told us to stop. Nothing else. No private address, no social-media activity, no inferred income, no list bought from a broker of personal data, no profile of anybody.
Where it came from. One of these, and we will tell you which one yours was if you ask: the organisation's own website, including the page the sentence in our message quotes; the public feed a podcast publishes, which carries the owner address its author chose to publish; a public video channel's own published contact address; a public page that linked to a tool like ours, which is how we found publishers who write about research tools; or a business-contact service we pay, which finds work addresses at a named organisation and holds no more about a person than a name, an employer and a role. For some addresses the record we kept is the page where we found the organisation rather than the page where we found the address, and for a few of the oldest it says only that the organisation was found: if yours is one of those we will say so plainly rather than guess, and delete it on request. We do not buy lists of consumers, and nothing here came from anything you did on our own site.
Why we think we may. Our legitimate interest in finding partners for a paid research tool, weighed against the interest of somebody receiving a message they did not ask for. What we do to keep that balance honest: we write to a work address about professional work the person published themselves; we send at most three messages and then stop whether or not anyone answers; we say who we are, where we are, and where we found you, in every one; one reply saying no, or one click, ends it for good and is recorded so that no later campaign can reach you. If you think the balance is wrong in your case, say so and we will delete the record rather than argue about it.
Where consent is required instead. Some countries require a person's agreement before a message of this kind, and do not accept a balancing of interests in its place. We do not send partnership messages in Germany or Austria at all for that reason. If you are reading this in a country whose law requires consent and a message reached you anyway, write to legal@valuescreener.ai and we will delete the record and tell you what was sent.
How to stop it, and how to see it. Three ways, all free and none of them needing an account: click the unsubscribe link in any message, which works at once and without signing in; reply with the word no, which a person reads; or write to legal@valuescreener.ai. Any of them stops every future message. The same address is where to ask for a copy of the record, a correction, or its deletion, and where to object to the processing itself under Article 21 of the GDPR. We answer within one month. You may also complain to the data protection authority where you live.
How long. Two years from the last message we sent you, then deleted. If you tell us to stop, we keep your address and nothing else, with no end date, because that entry is the only thing that can stop a later campaign reaching you by accident. Ask and we will delete the rest sooner.
Who else sees it. Two processors, both named in clause 6: the tool that sends the messages and holds the replies, and our customer system, which keeps the record of what was sent and what was answered. Neither is used for any email to a reader or a customer, and no prospect record is ever joined to a reader's account unless the same person opens one themselves.
15Automated processing
The nightly guardrail check is automated: it compares a figure to a threshold you chose and emails you when the comparison fails. Scores and rankings are computed by formula, and so is the plan that fits what you used, shown near the end of a trial. All of this produces information for you to read, not decisions about you. No output about a company is a decision about a person.
One decision about you is partly automated, and only in the direction that pays you. When you claim under the alerting commitment in the terms, our system reads our own records of what the nightly check found and what we sent you. Where those records show plainly that a guardrail broke and we then said nothing, it upholds the claim and refunds you without anyone reading it. It cannot reject a claim: every other outcome goes to a person, who decides it, and you may ask for a person in any case and contest what we decide. It is built that way so that the only thing a machine can do to you here is pay you, and so that a refusal is always a human one.
A refund of a first payment within thirty days is decided and issued by a person, from your request; no automated decision is made about it.
16Children
The service is for adults, and the terms require you to be 18. We do not knowingly collect personal data from children. If you believe a child has registered, write to legal@valuescreener.ai and we will delete the account and its data promptly.
17If something goes wrong
If personal data we hold is exposed or lost, we will investigate, contain it, and tell the competent supervisory authority within 72 hours of becoming aware where the law requires it. We will tell affected users directly, without undue delay, where the incident is likely to put them at risk, and we will say what happened, what data was involved, what we did, and what you should do. We would rather tell you early and imprecisely than late and neatly.
18Changes to this policy
When this policy changes materially we email your registered address at least 14 days before it takes effect, and we publish the new text here with a new version string. The version you accepted is recorded against your account, so it is always possible to establish which text applied at a given moment. Corrections that do not change your rights may take effect at once.