Skip to content

Privacy

Avenue Lift measures which version of a page converts better. It is built to do that without knowing who anyone is. This page lists exactly what is collected, what never is, and who handles it.

Last updated

Who this covers

Avenue Lift is a tool that website operators install to test different versions of their pages. Two groups of people are affected, and this page treats them separately:

  • Visitors of a website that has the Avenue Lift snippet installed. The operator of that website decides what is tested and is responsible for telling you about it; we process visitor data on their behalf and only to run their experiments.
  • Customers: people who sign in to Avenue Lift to run experiments.

What the snippet collects from visitors

The snippet has one job: remember which version of a page a browser was shown, and record whether that browser later completed a goal the site operator defined. To do that it collects:

  • A random visitor ID and a random session ID that the snippet generates in the browser. They are not derived from anything about you or your device.
  • Which experiments the visitor entered and which variant they were assigned.
  • Goal events the operator configured: a click on a chosen element, a page view, a form being submitted, a named custom event, or a purchase amount.
  • The URL of the page, without its fragment (the part after #).
  • Device type (desktop, tablet or mobile), whether the visitor is new or returning, the host name of the referring site, and the utm_source, utm_medium and utm_campaign parameters of the URL.
  • Country, taken from a header added by the hosting network. See what is never collected for how IP addresses are handled.

Site operators can send custom events. A custom event carries the name the operator chose and, optionally, one numeric value; any other property passed with it is discarded in the browser and never sent to us. Our terms require operators not to put personal data in event names.

What is never collected

  • IP addresses. The application does not read, store or log the visitor's IP address. Country comes from a header set at the network edge and nothing else about the address is kept.
  • Identity. No names, email addresses, account IDs or contents of form fields. A form goal records that a form was submitted, not what was typed into it.
  • Behavior recordings. No keystrokes, mouse movement, scroll maps or session replays.
  • Cross-site tracking. No third-party cookies and no identifier shared between different websites. A visitor ID exists only on the site that created it.
  • Fingerprints. The visitor ID is random. It is not computed from browser or device characteristics.

Cookies and browser storage

Everything the snippet stores is first-party: it is written under the domain of the website you are visiting and is not readable by other sites.

  • _lift_vid: the random visitor ID (cookie, mirrored in local storage).
  • _lift_sid: the random session ID (session storage; gone when the tab closes).
  • _lift_asg: which variant of each experiment this browser was assigned, so the page looks the same on the next visit.
  • _lift_consent: the consent decision, on sites that ask for consent first.

Clearing your browser's cookies and site data for a website removes all of these.

Do Not Track, Global Privacy Control and consent

By default, a browser that sends a Do Not Track or Global Privacy Control signal is left alone: the snippet stores nothing, sends nothing, and shows the original page. Requests carrying those signals are also dropped on the server. A site operator can turn this behavior off for their site; it is on unless they do.

Operators who need consent before any storage can run the snippet in opt-in mode. In that mode nothing is stored or sent, and no experiment runs, until the site's own consent banner tells the snippet that consent was granted.

Automated traffic (crawlers and other bots) is filtered out and not recorded.

How long visitor data is kept

The raw event log is deleted on a rolling basis; the default is 90 days. The per-visitor records an experiment's results are computed from (which variant a visitor ID saw, and whether it converted) are kept with the experiment so that its results stay reproducible.

Because none of this is linked to a name, an email address or an IP address, we cannot look up the records of a particular person. Clearing the site's cookies in your browser disconnects you from any existing records.

What we hold about customers

  • Account. Your name, email address and profile picture as provided by the sign-in service, and your role in each workspace. We do not store passwords; sign-in is handled by WorkOS.
  • Workspace content. The sites, goals, experiments, variants, brand notes and settings you create.
  • Activity. An audit log of changes made in the workspace (who did what, and when), and monthly usage counts for billing limits.
  • Billing. Your plan and subscription status. Card details are entered with Stripe and never reach our servers.
  • Email. We send product email to the address on your account: the weekly results digest, and alerts such as a winner being found, variants waiting for review or a usage limit being reached.

The marketing pages of this website do not set cookies and do not load third-party analytics. The only thing stored in your browser here is your light or dark theme preference, if you choose one.

What is sent to the AI

When a customer asks for test ideas, variants or a summary, Avenue Lift sends the following to Anthropic's API to generate the response:

  • The readable content of the page being tested, fetched from its public URL the same way a search engine would read it.
  • The brand notes the customer wrote for the site.
  • Earlier learnings for the site, the reasons reviewers gave when rejecting variants, and aggregate experiment results such as conversion rates.

Individual visitor records are not sent to the AI. By default, everything the AI writes for a running test is held for review and is not shown to a visitor until a person approves it. An admin can turn on auto-approve AI variants for a site; on that site, AI-written variants are shown to visitors without prior review.

Who processes data for us

We use a small number of service providers, each for one purpose:

  • Railway: hosting of the application and its database. Visitor and customer data is stored here.
  • WorkOS: sign-in and user accounts.
  • Anthropic: AI generation, with the inputs listed above.
  • Stripe: payments, when a workspace is on a paid plan.
  • Resend: delivery of product email.

We do not sell data and we do not share it with advertisers. A site operator can choose to forward experiment exposure events to their own analytics tool (for example GA4, PostHog or Segment); that happens in the visitor's browser, is off by default, and is governed by the operator's own privacy policy.

Your choices and requests

Visitors: turn on Global Privacy Control or Do Not Track in your browser to be left out of experiments on sites that honor it, or clear the site's cookies to reset your visitor ID. For anything else about a particular website, contact that website's operator: they control the experiments and the data.

Customers: you can change or remove your sites, experiments and members inside the product. How we secure all of this is described on the security page, and the rules for using the service are in the terms.

Changes to this page

When what we collect or who processes it changes, this page changes first and the date at the top is updated.