How it works
The short one is on the homepage. This is the version for people who want to know what the script records, how the AI reaches a judgement, what they can change, and what happens to their visitors' data.
One: it reads your site. Before it judges anybody, TribeMap crawls your pages and writes a description of your business. What you sell, who for, what your key pages are for, and what counts as somebody taking you seriously. That description is the thing everything else is measured against.
Two: the snippet watches visits. One script tag, about 5 KB, loading asynchronously. It records which pages someone saw, in what order, how long they stayed, how far they scrolled, what they clicked, and where they came from. It never reads what anybody types.
Three: an AI reads each visit as a story. Not a row of counters. It sees that somebody arrived from a LinkedIn ad, read two feature pages, opened pricing, left, and came back the next day straight to pricing. Then it decides how serious that was, and writes down why.
Four: it tells you what to do. Every morning you get a report: which marketing brought people who mattered, which brought nobody, and what to do about it.
Everything below is the detail underneath those four steps. If your question is not here, ask us in the chat.
No. TribeMap has no connection to your Google Analytics account and never reads from it. It has its own script and its own record of every visit, collected first-party on your domain. If you deleted Google Analytics tomorrow, nothing here would change.
No, and most people don't. They answer different questions: analytics counts what happened, TribeMap judges who mattered. Run both. We do.
Page paths in sequence with time on page. Clicks, scroll depth, and a count of typing activity, never its contents. Referrer, UTM parameters and ad click identifiers. Browser, operating system and screen size. An approximate city, derived from the IP address at the moment the event arrives.
It loads asynchronously and never blocks rendering: about 5 KB gzipped with zero third-party dependencies. The tag you paste is a small loader that pulls in TribeMap's own versioned SDK; beyond that it loads no third-party scripts, pixels or trackers, and it doesn't touch your DOM beyond inserting its own tag. If our service is unreachable, your site is unaffected.
Paste the snippet and give it a night. The first visits classify as they arrive; the first proper report needs a day of traffic behind it. On a quiet site, give it a week before you judge a channel.
Filtered in two places: in the browser, on headless and automation signals, and on our servers, on datacenter traffic. Bot sessions are kept separate rather than quietly inflating your numbers.
The shape of a visit, against what your business is for. Depth and order of pages, time spent, whether they reached the pages that matter, whether they came back, and what brought them. A visitor is sorted into one of three groups: interested, undecided, or gone.
A rules engine can't know that your /integrations page is where serious buyers go and your /careers page isn't. That's the difference: TribeMap reads your pages first, so the judgement is about your business rather than about generic thresholds someone picked.
Every classification stores its reasoning, so you can open a visitor and read why it decided what it decided. If the reasoning is wrong, it is usually wrong because the business context is wrong, and you can edit that yourself. See the next section.
Yes, and without work from you. It builds up claims about what your serious buyers do, tests them against new traffic, and drops the ones that stop holding. That is why the answers on day 30 are sharper than the ones on day 3.
This is the question people ask second and should probably ask first. The AI's understanding of your business is editable.
When you add a website, TribeMap crawls it and writes a business context: what you sell, who your customer is, what your important pages are for, and what counts as a real conversion. Open your domain's settings and you can read that description in full.
If it's wrong, you correct it. There is an override you can write yourself, and when it exists it is what the AI uses. Every judgement made afterwards is made against your version.
This matters more than it sounds. If your product page is full of jargon, if your real conversion is a demo booking rather than a signup, or if half your traffic is job applicants you don't care about, the context is where you say so, and the classifications change accordingly.
The rule of thumb: if the answers feel wrong, read the context before you blame the AI. It is nearly always the context.
No. There is no IP column anywhere in our database. The IP is used in memory at the moment an event arrives, for three things: deriving an approximate city from a local geolocation database, filtering datacenter traffic, and rate limiting. It is not sent to any third-party lookup service.
Typing is counted, never captured. The script does not read form field values, and password, email and search inputs are excluded from activity counting entirely. We collect no names, emails, phone numbers or addresses of your visitors.
One first-party cookie, scoped to your domain: a random ID so a returning visitor counts as one person rather than two. It expires after 12 months. No third-party cookies, no advertising cookies, and no cross-site tracking. If a visitor blocks cookies, a salted hash holds a single day together; the salt is rotated every 48 hours, so the hash can't be linked across rotations or reversed. Expired salts are then deleted by a daily cleanup once they have been expired for a further 72 hours.
For your visitors' data you are the controller and we are your processor, under a data processing agreement. We work to UK GDPR, EU GDPR and CCPA/CPRA. You are responsible for your own privacy notice and for any consent your visitors' jurisdictions require for the first-party cookie. Full detail is in the privacy policy.
Behavioural data only: page paths, sequence and timing, referrer and source, device type, approximate city. IP addresses are never sent to a model, and neither are names, emails or anything typed, because we never collect them. Providers are OpenAI and Google, and under both of their API terms the data is not used to train their models.
Yes. Deleting a website from your account deletes all of its visitor data. On account closure your data is deleted, apart from billing records we're legally required to keep. You can request an export or a deletion at any time at team@tribemap.ai. The security page has the rest.
It will classify every visit from day one, but a judgement about a channel needs enough visitors to mean anything. With a handful of visits a week you'll get useful individual dossiers and unreliable channel rankings. Read them as signals, not verdicts.
No, and it is not trying to. Visitors are pseudonymous behavioural profiles, not identified people. This is not a tool for putting a company name and an email against an anonymous visit.
No. If you need exact pageview counts, custom event pipelines or product analytics inside a logged-in app, keep the tool you have. TribeMap answers a narrower question: which of these people mattered, and what brought them.
Enough that it's a separate page: visitor intelligence vs web analytics, laid out side by side.
Paste one snippet. Give it a night.