Global Privacy Control analytics is a policy problem before it is a tooling problem. A visitor’s browser sends one persistent preference — do not sell or share my personal data — and your event pipeline has to decide, event by event, what it is still allowed to keep. Teams usually pick one of two shortcuts: ignore the signal, or drop every event from that browser. Both create a new problem while solving the old one.
This guide covers the event-level approach: read the signal where a page script cannot hide it, classify each attribute your analytics records, coarsen or drop what GPC constrains, and write down what you still collect for opted-out visitors. Your dashboards stay useful and your privacy notice stays accurate.
What GPC is, and what it is not
GPC is a browser-level preference, not a cookie and not a script. A supporting browser sends Sec-GPC: 1 on requests and exposes navigator.globalPrivacyControl === true to JavaScript. There is no banner to click and no storage to clear; the preference belongs to the browser and travels with it, including into private windows in supporting browsers.
It is an opt-out of sale and sharing, not a deletion request. California’s CCPA regulations name GPC as a signal businesses must honor; other jurisdictions cover it unevenly, so treat the legal scope as a question for counsel and the engineering as identical everywhere. Practically, GPC constrains cross-context identifiers and person-level profiling. It does not forbid counting page views.
Why page-level blocking breaks your numbers
Drop every event from a GPC browser and you remove a slice of traffic that is not random: privacy-conscious visitors cluster around specific browsers, regions and plan tiers. Funnel denominators shrink, conversion rates move, and retention cohorts shift for reasons that have nothing to do with your product.

Keep everything untouched and you retain the advertising identifiers and cross-site keys you implicitly told the browser you would not use.
Event-level decisions avoid both failures. Every event already carries attributes; the decision belongs per attribute class, not per visitor.
A three-tier event policy
Sort every field your analytics records into three tiers, implement the classification once in your tracker, and re-apply it on the server.
| Tier | Rule | Typical fields |
|---|---|---|
| Drop | Identifiers and cross-context linkage | ad click IDs, third-party identifiers, cross-site stitching keys, profile joins, session replay for that session |
| Coarsen | Keep the count, remove the person | page path with the query string stripped, hour-bucketed timestamp, browser family, region instead of IP |
| Keep | Not tied to a browser identity | server-side counters, error logs with identifiers removed, aggregate totals |
Two details matter more than the table. First, when you coarsen, generate a fresh random nonce per request instead of reusing a visitor identifier — otherwise you have rebuilt the profile you just rejected. Second, run the same function server-side, because a client can be modified.
// one function, both edges
function tierFor(field, gpc) {
if (!gpc) return 'keep';
if (IDENTIFIER_FIELDS.has(field)) return 'drop';
if (CONTEXT_FIELDS.has(field)) return 'coarsen';
return 'keep';
}
Read the signal on the server, and keep the decision
Read Sec-GPC from the request headers in your edge or API layer, and use navigator.globalPrivacyControl only as a secondary value for the same request. If you rely on client-side detection alone, a blocked script hides the preference and you silently record everything.

Then store the decision, not just the signal: the value, its source, a timestamp, the identity you keyed it on (hashed) and the policy version you applied. Compare header and client values and log the mismatch — a fabricated signal only costs the sender data, but a suppressed one costs you accuracy.
What you still record for opted-out users
Write this list into your privacy notice before anyone asks for it:
- Event names and counts — kept. This is what keeps funnels and usage metrics honest.
- Page path — kept, with the query string removed; URLs routinely carry email addresses and campaign parameters.
- IP address — not stored; resolve it to a region at the edge and discard it.
- Full user agent — reduced to browser family and major version.
- Session replay and heatmaps — off for that session unless the visitor opts in through your own interface.
- Logged-in product analytics — a separate basis, not a GPC exemption. Keep those tables apart and document which basis each one relies on.
Publish the same description in your customer-facing terms and keep it aligned with your processing agreement, for example pryntid.com/dpa. A privacy-first setup is only auditable when the retention and purposes you signed match what the pipeline actually does.
Where Prynt Analytics fits
Prynt Analytics covers 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. Excluding bots keeps your denominators honest while the GPC gate keeps them defensible, and both live in the same workspace as Identify, Control and Replay.

FAQ
Does honoring GPC mean I lose all analytics for opted-out visitors?
No. You lose identifiers and cross-context linkage, not the count. Coarsened events keep usage totals and funnel denominators; what you cannot do is attribute them to a person or join them across sites.
How do I know a GPC signal is genuine?
Read the Sec-GPC header server-side and compare it with navigator.globalPrivacyControl for the same request. A fabricated signal only reduces what you collect, so the real risk is suppression — which is exactly why the mismatch gets logged.
Can opted-out visitors stay in my conversion funnel?
Yes, if the funnel step is recorded without a stable identifier. Decide once whether coarsened events enter funnels, write the rule down, and apply it to every report; two dashboards with different rules will disagree and neither will be trusted.
Try it free
Prynt Analytics counts what you need without profiling the people who asked you not to. Start the 14-day trial at pryntid.com/platform and read the implementation notes in the docs.
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.