Skip to content
Privatus Analytics

Privacy

Cookieless web analytics that measures without tracking people

Every privacy claim we make has a technical explanation. This page describes exactly what happens to a pageview, what we store, what we never store, and how you can check it yourself.

Don’t take our word for it

These tools are run by other organizations. Each link scans this website fresh, so you see today’s result, not a screenshot we chose.

The data journey

Every visitor request, including requests from the EU, EEA, UK and Switzerland, is processed in our US region. The IP address and user agent are used in memory while the request is handled, and dropped before the minimized event is stored.

  1. Step 1 · Anywhere
    Visitor’s browser
    • Page URL and referrer
    • UTM parameters
    • Screen width and language
    • Your custom event properties
    • No cookies, and no local storage except a visitor's own opt-out choice
    • No fingerprinting APIs
    • No third-party requests
  2. Step 2 · United States
    Ingest, in memory
    • Country from the Cloudflare header
    • User agent → browser, OS, device
    • Bot checks and your traffic rules
    • In-memory visit key
    • IP address dropped after the request
    • Full user agent dropped after the request
    • PII scrubbed, query strings trimmed
  3. Step 3 · Queued in the US region
    Minimized event
    • The minimized event only
    • No IP address
    • No raw user agent
    • No visit key or salt
  4. Step 4 · United States
    Storage and dashboard
    • Pages, sources, campaigns
    • Country and device categories
    • Random visit ID (one day at most)
    • Event properties
    • Nothing that identifies a person
    • Nothing that links days, sites or devices

What we collect, and what we never collect

Collected (after minimization)

  • Page. Hostname and path. Query strings removed except campaign parameters and your allowlist. Fragment removed unless you use hash routing.
  • Referrer. Domain and path of the referring page. Query removed, internal referrers blanked.
  • Campaign. UTM parameters and ref. Ad click IDs (gclid, msclkid, fbclid) are reduced to a yes/no flag.
  • Location. Country only, from the country header our network provider (Cloudflare) adds to each request. No region, city or coordinates.
  • Technology. Browser, operating system (family and major version), device type, screen size range, language.
  • Engagement. Visible time on page and scroll depth in 10% bands.
  • Time zone. The name of the browser’s time zone, such as America/Denver. A browser running on a server is rarely set to the time zone of the country its request comes from, so we compare the two to help tell bots from people. That happens in memory, then the time zone is dropped. It is never used as a location and never stored with a visit or an event. It is kept only in the bot training records of Advanced Bot Detection (honeypot hits, and samples from our own sites) for 12 months.
  • Events. Names and up to 30 typed properties you choose to send, after PII scrubbing.

Never collected or stored

  • IP addresses. Used in memory for bot checks and the daily visit key, then dropped. Never written to disk, logs, queues or databases, not even for bots.
  • Full user agents. Parsed into categories in memory, then dropped.
  • Cookies and device storage. No cookies, localStorage, sessionStorage or IndexedDB (except a visitor’s own opt-out choice).
  • Fingerprints. No canvas, WebGL, font, audio or battery APIs, and no extra client hints.
  • Persistent identifiers. Nothing that lasts beyond one day, or links sites or devices. No user profiles.
  • Precise location. No coordinates.
  • Third-party requests. The tracker talks only to Privatus Analytics.

PII scrubber. Before anything is stored, paths, referrers and event properties are checked for email addresses, long identifiers, token-like strings and long digit runs such as phone or card numbers. Matches are replaced with placeholders like [redacted-email], and each site sees how many redactions happened so leaks can be fixed at the source.

How visits are counted without cookies

  1. Each UTC day we generate a random 256-bit salt. It exists only in memory and is destroyed at the end of the day, plus a 30-minute grace period.
  2. For each hit we compute HMAC(daily salt, site, IP, user agent) in memory and look it up in a memory-only store with a 30-minute inactivity timeout, capped at midnight UTC.
  3. A new visit gets a random visit ID that is not derived from that key. Only the random ID is stored.
  4. When the salt is destroyed, nothing can link stored visits back to a device or to each other across days.

This is pseudonymous data: heavily minimized and not linkable beyond one day. We don’t describe it as anonymous.

Every field we store

Adding a stored field requires a written review of its purpose, minimization and retention. This is the current list.

FieldPurposeMinimizationRetention
Hostname, pathPagesQuery stripped except allowlist, PII scrubbed, path masksPlan
Referrer host and pathSourceswww. and m. stripped, query dropped, internal blankedPlan
UTM parameters, ref, paid clickCampaignsClick IDs reduced to a yes/no flagPlan
CountryLocationFrom the Cloudflare country header, IP droppedPlan
Device, browser, OS (with major versions), screen rangeTechnologyRaw user agent dropped, screen bucketedPlan
Visit IDVisitsRandom UUID, one UTC day at mostPlan
Event propertiesEventsUp to 30 keys, typed, PII scrubbedPlan
Time zone nameTraining the bot scoreNever stored with a visit or an event. Kept only with a honeypot hit or a sample from our own sites12 months
Filter countersBots, rules, redactionsAggregates only, no IP13 months
Visit key (memory store)Visit continuityHMAC key, 30-minute timeoutOne day at most
Daily salt (memory store)Visit keyNever persisted1 day + 30 minutes

“Plan” retention: 3 months on Free, unlimited while subscribed on paid plans, with per-site overrides.

Verify it yourself

  1. Open this page’s developer tools and go to the Application (or Storage) tab. You’ll find no cookies, and nothing in local storage from the tracker. The only thing the tracker ever stores is a visitor’s own opt-out choice. This site also keeps your light or dark theme there if you change it.
  2. Switch to the Network tab, filter by /api/event and reload. You’ll see one small request per pageview.
  3. Open the request payload. It contains only the fields listed below: no identifiers, no fingerprint.
  4. Add data-debug="true" to the script tag on your own site and watch the Console. Every hit is logged with the prefix [privatus], exactly as it would be sent, and nothing leaves the browser. A hit that is skipped logs the reason, and it works on localhost. See Test with nothing sent.
  5. Read the tracker’s source. It’s MIT-licensed and short enough to read in one sitting. See open source.

Try the cookie and tracker scanner on your own site, too.

What a request contains

These examples show every field that can appear in each kind of hit. Most hits carry far fewer. The table below says when each field is sent.

A pageview
{
  "s": "pa_7Q2K9XH3AB",
  "t": "pageview",
  "u": "https://example.com/app?utm_source=newsletter#/pricing",
  "l": "en-US",
  "w": 1440,
  "i": 0,
  "r": "https://news.example.org/",
  "h": 1,
  "d": 1,
  "g": 1,
  "hp": 1,
  "a": "webdriver",
  "tz": "America/Denver",
  "xp": "var_4M8T2KQ9ZD",
  "p": { "plan": "pro" }
}
An engagement hit
{
  "s": "pa_7Q2K9XH3AB",
  "t": "engagement",
  "u": "https://example.com/app?utm_source=newsletter#/pricing",
  "l": "en-US",
  "w": 1440,
  "i": 1,
  "r": "https://news.example.org/",
  "h": 1,
  "d": 1,
  "g": 1,
  "hp": 1,
  "a": "webdriver",
  "tz": "America/Denver",
  "e": 48210,
  "sd": 80
}
A custom event
{
  "s": "pa_7Q2K9XH3AB",
  "t": "event",
  "u": "https://example.com/app?utm_source=newsletter#/pricing",
  "l": "en-US",
  "w": 1440,
  "i": 1,
  "r": "https://news.example.org/",
  "h": 1,
  "d": 1,
  "g": 1,
  "hp": 1,
  "a": "webdriver",
  "tz": "America/Denver",
  "n": "signup",
  "p": { "plan": "pro" }
}
FieldMeaningWhen it is sentWhat happens to it on our side
sSite IDEvery hitUsed to find the site. Stored as the site the event belongs to.
tHit type: pageview, engagement or event. The optional Web Vitals and click map modules add vital and click.Every hitStored as the kind of event.
uPage URL, with non-campaign query parameters already removed in the browserEvery hitHostname and path are stored after path masks and the PII scrubber. Campaign parameters are stored, ad click IDs become a yes/no flag, and the rest of the query is dropped.
lBrowser languageEvery hitStored, cut at 16 characters.
wScreen width in pixelsEvery hitStored only as a size range. Read in memory for the bot score. The exact width is dropped.
i1 after the first real input on the page (pointer, key, touch or wheel), otherwise 0Every hitStored as a yes or no on the visit. Nothing about the input itself is sent.
rReferrerOnly when there is oneHost and path are stored after the PII scrubber. The query is dropped, and a referrer from your own site is blanked.
h1 when hash routing is usedOnly when the fragment is part of the page addressUsed in memory to keep the fragment as part of the path. Not stored.
d1 when the browser sends Do Not TrackOnly when it is onIf the site honors Do Not Track (off by default), the hit is dropped and nothing is stored. Otherwise it is ignored. The flag is never stored.
g1 when the browser sends Global Privacy ControlOnly when it is onIf the site honors Global Privacy Control (on by default), the hit is dropped and nothing is stored. Otherwise it is ignored. The flag is never stored.
hp1 on a page the site owner marked as a honeypot, 0 on a page marked as reached by people onlyOnly on pages marked for Advanced Bot DetectionA hit with 1 never becomes a visit or an event. It is counted as a bot and, on plans with Advanced Bot Detection, kept as a honeypot hit. A hit with 0 is counted as usual, and only our own sites keep a copy as a training sample.
ch1 on a "confirm you're human" page that a site shows a random few visitors, people and bots alikeOnly on pages marked data-hp="ch"A hit with 1 never becomes a visit or an event and is not counted as a bot. On plans with Advanced Bot Detection it is kept next to the honeypot hits, marked as a challenge.
aAutomation name: webdriver, phantom, nightmare or cypressOnly on a honeypot or challenge page, when an automated browser is detectedKept with the honeypot hit. Ignored everywhere else.
tzTime zone name, such as America/DenverEvery hit, when the browser reports oneCompared in memory with the country of the request for the bot score, then dropped. Never stored with a visit or an event and never used as a location. Kept only with a honeypot hit or a training sample.
xpExperiment preview value, from pa_preview in the addressPageviews only, when the address has pa_previewUsed in memory to show the previewed version. The pageview is not counted for the experiment. Not stored.
pCustom propertiesPageviews and custom eventsUp to 30 are stored after the PII scrubber. Revenue and currency are stored as their own fields.
eVisible time on the page, in millisecondsEngagement hitsStored, capped at 4 hours.
sdScroll depth, in percentEngagement hitsStored in 10% bands.
nEvent nameCustom eventsStored, cut at 120 characters.
vWeb Vital valueVital hits (Web Vitals module)Stored with the name of the vital, for a sample of hits.
cShort CSS selector of the clicked element, or the name the site gave itClick hits (click map module)Stored, cut at 512 characters. No coordinates or text.

The optional survey module sends its own small requests. They are listed with the full reference in the docs. What a honeypot hit keeps is described in Advanced Bot Detection.

Where your data lives

US storage and processing, and transfer safeguards.

Compliance hub

GDPR, ePrivacy, US state laws and more.

Data Processing Agreement

Included in our terms.

Start measuring without cookies

No cookies, no fingerprinting and no consent banner needed for measurement.

Your first workspace starts with 14 days of Pro, free. After that it moves to the free plan, with 10k events a month, 3 sites and 3 months of history. No card required.