Most product teams still learn their funnel from a cookie. When the banner defaults to “reject,” or the browser sends Global Privacy Control, that funnel loses the step that used to be silent: the identifier that turns five page views into one visit and three visits into one user. Cookieless product analytics replaces that assumption with an explicit one — you decide, signal by signal, what you measure before consent and what waits.
The trap is treating “cookieless” as a compliance result. It is an architecture. Page paths and session-scoped counters can be collected without an identifier; custom events with properties, user identification, cross-session retention and session replay cannot, unless you have a lawful basis and a record of it. Strip everything and your funnel goes blind after the landing page. Collect everything and your consent log becomes fiction.
Here is where the line falls, how to honor Global Privacy Control as a machine signal, and how to plan funnels that stay readable across two tiers.
Split measurement into two tiers before you touch the tracker
Write the split down first — a policy document, not a tracker checkbox — short enough that an engineer can implement it from memory.
Tier one: page path, no identifier
Allowed in most configurations: route, referrer host, timestamp, device class and viewport, coarse country. Nothing here survives the tab. Hold the session id in memory, not in a cookie or localStorage, and drop it when the tab closes. That is enough to count visits, entry pages and drop-off between pages.
Tier two: anything that names a user or a conversion
Custom events with properties, identify(userId, { accountId }), cross-session retention, experiments and replay all depend on linking a person or device across time. They wait for consent and carry the consent state with them. Then check the list against your privacy notice: the processing described there has to match what the code does.
Honor Global Privacy Control as a signal, not a banner
Global Privacy Control is a browser-level opt-out: Sec-GPC: 1 on the request, navigator.globalPrivacyControl in the page. California’s privacy law makes it binding for covered businesses as an opt-out of sale and sharing. Two rules follow.

First, a visitor who sends it has opted out: do not set an identifier, do not call identify(). Record the signal on the session and keep the session in tier one.
Second, it is not a replacement for your consent flow. If your consent tool collected a rejection, GPC does not turn it into consent, and a later banner click should not silently override the signal. Log which of the two decided, so the record answers the question months later. Where your own obligations end is a question for counsel — have counsel review the mapping before you ship.
Name events so a consent flag is enough
The cheapest way to keep a two-tier dataset readable is naming discipline.
- One name per action, both tiers. Use
signup_completedbefore and after consent; only the properties and the identifier change. - Keep page views dumb. A
page_viewcarries the path and nothing personal, so it can always be collected. - Put consent on the event, not beside it. A flag and a timestamp let you filter later without re-instrumenting.
- Never put an email or phone number in a property. Masking protects form fields; it does not protect properties you send by hand.
Version the dictionary in one place — our docs cover event and property setup at https://pryntid.com/docs.
Plan funnels around a visible boundary

Use page-path steps as the denominator
A top of funnel built from page paths — landing, pricing, signup form view — runs on every session, consented or not. That is your denominator, and it tells you whether the page is doing its job.
Stop consented funnels at the wall and label them
Activation and revenue steps exist only for consented users. Name the step accordingly: “signup (consented sessions)”.
Do not extrapolate silently
The consented funnel describes consented sessions only. You can compare the two tiers on page-path steps to see whether consenting visitors behave differently — that is a real signal. Do not scale a consented conversion rate onto the whole population and call it a funnel.
What identity looks like without cookies
A device fingerprint returns a visitorId without writing storage, which makes it tempting to treat as a consent-free pass. It is not one: fingerprinting is still processing tied to a device. The common split is purpose — use it for security decisions such as bot, AI-agent and VPN detection, and keep product analytics in tier one until consent arrives. Our device fingerprinting page explains how the signals are built: https://pryntid.com/device-fingerprinting.
Where Prynt Analytics fits
Prynt Analytics is privacy-first product analytics: overview, live view, custom events, funnels of two to eight steps, goals, weekly retention and paths, with bot sessions excluded and Global Privacy Control honored. It runs at $39/month after a 14-day trial. Data processing terms are at https://pryntid.com/dpa.

FAQ
Is cookieless analytics automatically consent-free?
No. Cookieless describes the storage mechanism. Whether consent is required depends on what you measure and where the visitor is, and counting page paths is a different act from identifying a user across visits.
Does Global Privacy Control replace our consent banner?
No. It is an additional opt-out you must not overwrite. Keep both, and log which one decided.
Can we measure retention without consent?
Within a session, yes. Across sessions, no: retention needs an identifier, so it belongs in tier two.
Try it free
Start a 14-day trial at https://pryntid.com, send page paths without an identifier from day one, and turn on custom events when consent arrives.
Try it free
Prynt is device intelligence with a free plan: 5,000 identifications a month, visitor IDs with confidence and bot detection — no credit card. Upgrade to Pro for VPN, proxy, tampering and behavioral signals.