Both tools lose data when a visitor declines cookies, and they lose different things. This is what GA4's consent mode and Piwik PRO's anonymous mode each record, what is observed and what is estimated, where the EU's cookie reform stands, and a test to run on one site.
In Sweden a site has to ask before it sets an analytics cookie, and saying no has to be as easy as saying yes. So some of your visitors say no, and the question is what your analytics tool still knows about them. GA4 and Piwik PRO give different answers, and the difference decides how far you can trust a funnel or a conversion rate.
This is a design to test, built from the two vendors’ own documentation. It reports no run, so it is not a benchmark, and it is not legal advice: whether a setup is allowed is a question for a lawyer.
After a no, a visit ends up in one of three states. It can be recorded thinly, meaning the tool saw it but without an identity. It can be modeled, meaning the tool estimates it from similar visitors who said yes. Or it can be missing, meaning nothing was collected. Which state you get depends on the tool, on how you set it up and, for modeling, on your traffic.
The Swedish rule is the reason this matters. PTS, the authority that supervises it, says a site must ask for consent before it sets cookies, that collecting statistics is not a “necessary” cookie, and that the buttons for yes and no must be designed alike. So some visitors will always say no, whichever tool you pick. Missing visits hurt totals more than rates: revenue in the tool undercounts by the share of buyers who declined, so totals come from the store, as in the reconciliation article. Rates and funnels can be trusted only if people who say no behave like people who say yes, and you cannot check that from the tool alone.
GA4 depends on how you set up consent mode. In basic mode the Google tags do not load until the visitor chooses, so after a no nothing is sent, not even the consent status. In advanced mode the tags load with consent denied by default, and while it is denied they send measurements without cookies: a timestamp, the user agent, the referrer, the consent state and a random number per page load.
On top of that GA4 can model. Google’s behavioral modeling estimates the behavior of visitors who decline from similar visitors who accept. It is switched on only when the property qualifies:
Meeting these does not guarantee it, and when the model lacks confidence the events from visitors who declined are simply not reported. Modeled data shows only in the Blended reporting identity, and some reports, such as retention and segments with sequences, leave it out. A store with fewer than a thousand consenting visitors a day does not meet the second condition. Take a made-up store with 600 visitors a day, 60 percent of whom accept. That is 360 consenting visitors, far short of 1,000, however many events the other 240 generate. GA4 will not model them, so for that store a no is a missing visit.
Piwik PRO is a suite: analytics, a tag manager, a consent manager and a data activation layer. It can be hosted in a private cloud, in the public cloud in the US, the Netherlands, Germany or Hong Kong, and in EU-operated data centers in Sweden.
After a no, Piwik PRO lets you choose between collecting nothing and collecting anonymous data. In anonymous mode the IP address is fully masked, you know only the country, and you cannot tell a new visitor from a returning one. To tie the events of one visit together you can use a session hash, built from the IP address, operating system, browser name, version and language, plugins and site ID, or two cookies that expire 30 minutes after the last event, or both. Attribution for these visitors is last click at most. Piwik PRO’s documentation describes no modeling: what is collected is counted as it is. One trap it names: some Tag Manager triggers, such as returning visitor, traffic source or firing a tag once per session, set cookies of their own, so tags that use them need the right consent type.
Piwik PRO’s own help page is careful about the legal side. Whether you may collect anything from visitors who do not consent, it says, depends on privacy laws and how your legal team interprets them, and you should tell visitors that you do it. Its Consent Manager can also pass the consent state to Google tags through consent mode v2, so one banner can drive both tools. It documents how to connect a third-party consent tool too (Cookiebot has a ready integration), but that integration does not pass on consent statistics, so the acceptance rate has to come from the consent tool.
| After a no | GA4 | Piwik PRO |
|---|---|---|
| Record nothing | Basic mode: tags blocked, nothing sent | Collect nothing |
| Record a thin version | Advanced mode: cookieless pings with consent state, referrer and user agent | Anonymous mode: IP address masked, country only, a session hash and/or 30-minute cookies tie one visit together |
| Estimates | Behavioral modeling, if the thresholds are met | None |
| What stays missing | Returning visitors, and everything below the thresholds | New against returning, and attribution beyond last click |
| Hosting | Run by Google | Your choice of cloud, including EU-operated data centers in Sweden |
| Bundled | Analytics only; Tag Manager is a separate product | Analytics, tag manager and consent manager in one suite |
I would not choose from tables. I would run both on a site of my own, or a test store, never a client’s, behind one consent banner. Piwik PRO’s Consent Manager, or another consent tool, holds the choice and passes it to the Google tags, with GA4 in advanced mode. Whether to switch Piwik PRO’s anonymous mode on, and whether GA4’s advanced mode is acceptable on the site, are legal decisions, so I would take them before the test, not after.
Each day I would write down the store’s orders, the purchases GA4 shows in the Blended and in the Observed reporting identity (switching between them changes no data), the purchases Piwik PRO shows for consenting and for anonymous visitors, the sessions in each tool and the share of visitors who accept the banner, taken from the consent tool. Then I would divide each tool’s purchases by the store’s orders, the way the reconciliation article does. Three things to look for:
It will be one site, so it is a trial and not a benchmark, and I would say so wherever I quoted it.
| If | I would lean to |
|---|---|
| Most paid spend is on Google, and its bidding should have modeled conversions | GA4 in advanced mode, since consent mode feeds Google’s own modeling |
| Where the data lives, and fewer estimates, matter more than modeling | Piwik PRO |
| Traffic is low | Expect modeling to stay off, so compare what each records for a no |
| You need both | One consent source, both tools, and a person who owns the setup |
On November 19, 2025 the Commission proposed moving the cookie rules into the GDPR. Its Article 88a would allow storing or reading data on a device without consent for aggregated information that measures the audience of an online service, when the controller of that service does it “solely for its own use”. As of Parliament’s tracker on September 20, the proposal is still in committee. The rapporteurs published a draft report on June 22 and left the cookie provisions to the committees, more than 1,750 amendments to the draft were tabled, and the Council vote on its position, planned for June 26, was canceled. Whether either of these setups would fit that wording is a question for a lawyer.
So I would decide under today’s rules and build so that a change is a setting, not a rebuild: consent in one place, every tag reading it.
Neither tool makes a no free. GA4 can estimate what it cannot see, but only above a threshold many stores will not reach. Piwik PRO can record a thin anonymous version of every visit, if your lawyer agrees. The honest comparison is a few weeks of both on one site, set against the store’s own orders.
Sources: Google, Consent mode overview and Behavioral modeling for consent mode. Piwik PRO, Can I collect any data from visitors who don’t consent?, Google consent mode v2 integration and Privacy compliance. PTS, Kakor (cookies). European Commission, Digital Omnibus proposal, COM(2025) 837, and European Parliament, Legislative Train: Digital Omnibus (updated September 20, 2026). These rules and settings change often, so some details here may have moved on.
ChatGPT, Gemini and Copilot recommend products now, and the sale happens on your store. This is what they read in a feed, which of the new programs reach a Nordic store yet, and how to see their visits in GA4 without fooling yourself.
Google now ships official MCP servers for Analytics, Ads and BigQuery, so an agent can answer questions about your numbers. This is how I would give it a seat instead of the keys: read-only identities, ten questions with known answers, and the ways it still goes wrong.
Meta, GA4 and a store can all be right and still disagree. This is what each one counts after Meta's March 2026 attribution change, how to split the gap into named causes, and which number to trust for which decision.