On 12 March 2026, France's data protection authority, the CNIL, adopted a recommendation on tracking pixels in emails (deliberation 2026-042). It was published in the Journal officiel on 14 April 2026, and a question-and-answer page followed on 22 July 2026. The recommendation does not create a new law. It spells out how an existing rule, article 82 of the French data protection act, applies to the pixels that email tools use to count opens. Tracked links fall under the same article, although the recommendation does not cover them directly.
This article is for the people who decide how a SaaS measures its email: the DPO, the head of marketing, and the CTO who chose the email stack. It covers what changes for a SaaS's emails, what the rules and their exemptions say, how to measure a campaign without relying on opens, and how Notifizz applies consent to measurement.
What changes for your SaaS emails
Take a SaaS that sends three kinds of email to people in France: receipts, password resets, and an onboarding sequence. Its email tool probably counted who opened each of them. Under the recommendation, counting a person's opens to judge or improve a campaign requires that person's consent, whatever the email. A receipt or a password reset can carry an exempt pixel only for two narrow purposes, the security of an authentication and deliverability. The onboarding sequence has no exemption for campaign measurement.
The recommendation applies whatever kind of organisation sends the email, and whatever the recipient's status: B2B email and employees' work addresses are included (FAQ, question 3). In short:
- Sending the email is a separate question. The recommendation is about the pixel, and consent to the pixel can be needed for an email that needs no consent to be sent.
- Measuring a person's opens to analyse or optimise a campaign requires their prior consent (section 3.1).
- The security of an authentication is exempt, within the conditions of section 3.2.
- Deliverability is exempt only to adapt or stop sends to inactive recipients, keeping in principle the date of the last open, by day, and only for emails the recipient asked for or that relate to a service they asked for (section 3.2, FAQ question 5).
- Aggregate statistics can reuse data collected under an exemption, once it has been effectively anonymised (FAQ question 6).
- Tracked links are not covered directly by the recommendation, but they remain trackers under article 82 (FAQ question 1).
- A withdrawal means that pixels already placed are no longer exploited, and that data with no other legal basis is deleted (section 5, FAQ question 22).
What the CNIL's recommendation says about tracking pixels
A tracking pixel is a one-pixel transparent image loaded from the sender's server when the email is opened. It tells the sender which recipient opened the email, and when. Most email tools use it to calculate an open rate.
Under article 82 of the French data protection act, a tracking pixel falls under the same rules as cookies and other trackers. The recommendation follows the EDPB's Guidelines 2/2023, which treat email tracking pixels as access to the recipient's terminal equipment under article 5(3) of the ePrivacy directive.
Consent is required when the pixel is used to analyse open rates in order to measure and optimise campaign performance. That is the first purpose listed in section 3.1, and it is the core use of an email tool's statistics dashboard.
Section 3.2 sets out two exemptions. The first is the security of an authentication, for instance checking that an email carrying a login code is opened on a device known to belong to the user. The second is deliverability, strictly limited to adapting the frequency of sends to inactive recipients, or stopping them. Under that exemption, in principle only the date of the last known open is kept, by day, without the hour, and only for emails the recipient asked for, or that relate to a service they asked for, such as transactional emails.
The exemptions do not allow placing a pixel to follow each person's opens for campaign performance. They do not forbid every statistic either: question 6 of the FAQ says that data lawfully collected through an exempt pixel can be reused for open statistics without further consent, once it has been effectively anonymised, for instance by aggregation. Anonymising at collection is not a way around consent, though: question 18 recalls that article 82 applies whether the data collected is personal or not.
Transactional emails are not a blanket exemption
A common misreading of the recommendation is that transactional emails are exempt from the consent requirement. They are not.
The recommendation includes a "Point d'attention" stating that consent to a tracking pixel can be needed for an email that does not itself need consent to be sent. An order confirmation is the example given. Question 16 of the FAQ adds that a purchase confirmation can fall under the deliverability exemption, while an abandoned-cart reminder cannot.
The boundary is not the email's category. It is the purpose of the tracking. If the pixel is used to measure a campaign's performance, to calculate an open rate, or to feed an attribution funnel, consent is required, even in a transactional email.
For a SaaS, a pixel in a password reset, a receipt or a security alert can therefore only be exempt within those two narrow purposes. Tracking which recipients opened the receipt, or testing a reworded reset email on each person's opens, requires consent.
Tracked links, withdrawal, and the transition period
The recommendation is about pixels, and question 1 of its FAQ says it does not cover tracked links directly. They still fall under article 82: a link that carries an identifier and logs the click before sending the recipient on is a tracker, and whether it needs consent depends on what the click is used for. The recommendation names one kind of exempt tracked link: a unique link per recipient that protects a feature reserved to them, such as the link to withdraw consent, which counts as a security measure. Click tracking used to measure a campaign has no such exemption.
A withdrawal has to take effect on emails already sent. Section 5 asks for solutions so that trackers already placed are no longer exploited, notably when the recipient opens an older email again. Question 22 adds that the data collected before the withdrawal is deleted when no other legal basis justifies keeping it. And under article 7(1) of the GDPR, the controller must be able to demonstrate that each person consented.
For addresses collected before 14 April 2026, senders could keep using pixels after giving recipients clear information, unless they objected, within a period that should not in principle exceed three months, and that ended in mid-July 2026. The FAQ allows that period to be extended reasonably where difficulties are objectively justified and documented (question 25). A sender that never sent that information is expected to ask for consent, or to stop the pixels that need it (question 26).
How to measure a campaign without relying on opens
A lower measured open rate under consent does not mean lower engagement. It means fewer people are observed. The people who refused measurement may have opened the email, clicked the link and converted. Open and click statistics just do not see them, while a conversion on a business event can still be counted, as described below. Each indicator depends on something different:
- Opens depend on a pixel. Their limits: consent, and images loaded automatically.
- Tracked clicks need no pixel, but the tracked link is itself a tracker. Their limits: consent, depending on what the click is used for, and links followed by robots.
- Delivery and bounces depend on no tracker. Their limit: they do not prove the email was read.
- Business events depend on no email tracker. Their limit: they need instrumentation in the product.
Three approaches follow from that.
First, measure outcomes on business events. Decide which event in your product means the email worked: an invoice paid, a feature activated, a trial upgraded. Those events are tracked by the product, not by an email pixel, so measurement consent does not decide whether they are counted. Their processing still has to respect the GDPR on its own terms, like any product data. In Notifizz, a campaign declares the event that counts as its objective: a click, a read, or any custom event the customer's backend tracks. When the campaign is set to end on its objective, a recipient who reaches it gets none of the remaining messages. A conversion on an event the customer's backend sends counts whether or not the person is measured, while click and read objectives depend on measurement.
Second, read the delivery side. Delivery does not depend on a pixel. In Notifizz, a campaign that shows "Not measured" on opens still shows how many emails were delivered and how many bounced, and its send attrition counts every recipient who was dropped, with the reason, such as an invalid address, a bounce, a complaint or an unsubscribe. Delivery history is kept in every state.
Third, read a partial figure for what it is. Where nothing could be measured, Notifizz's statistics screens show "Not measured", never 0%. Where only some recipients are measured, the figure is marked "partial": it counts the opens and clicks of measured recipients only. A partial figure is therefore a floor of the open signals observed, not a count of readers. It can understate engagement, because recipients who were not measured are not in it. It can also overstate it, because opens are a noisy statistic for every email tool: mail clients and security gateways load images and follow links before anyone reads, and Notifizz does not yet tell a security gateway's click apart from a person's. Fifty measured opens mean fifty open signals, not fifty people who read the email.
How Notifizz applies consent to measurement
The customer, as controller, collects consent. Notifizz applies it to every email, closed by default. The measurement consent documentation covers the setup in detail; here is how it works.
Each organisation declares, in its dashboard, how measurement applies to the people who have given no answer. Until it declares, nobody is measured. The declaration is a certification made by a signed-in member of the customer's team, never by Notifizz staff, and every answer is appended to a journal with who, when, and the version of the wording that was on screen. That journal records the organisation's declaration. It does not replace the evidence of each person's consent, which the customer must be able to produce.
Person by person, the customer's backend passes on a consent or a refusal with one call to the server SDKs, setMeasurementConsent(), made once when the answer changes, not on every send. For the same person, the latest answer replaces the previous one, so a person who changes their mind can be measured again. When an email goes to one of the customer's users at a given address, and a refusal was passed on for that address while a consent was passed on for the user id, the refusal wins: a send designates both. A refusal the person expressed from an email prevails over any consent the customer passes on, until the person lifts it.
Consent is read at the moment of the open or the click, not only when the email left, so a refusal also stops the counting on emails already sent. An email that is not measured carries no pixel, and its links point straight at their destination. Refusing measurement never removes anyone from a send.
Every promotional email, and every transactional email whose opens and clicks are measured, links to a hosted preference page. There, a person can stop the measurement of their emails while still receiving them, or unsubscribe from promotional email, which also stops the measurement. Messages about their account, such as receipts and security alerts, keep arriving, unmeasured. Notifizz records these choices on the customer's behalf.
What Notifizz does not do
Notifizz does not make a customer compliant, and does not give legal advice. The customer is the controller: it decides how consent is asked, and must be able to prove it. Notifizz applies the answers it receives, and records the choices people make on the pages it hosts.
Notifizz does not tell a security gateway's click apart from a person's. It has no per-link report. A single link cannot be kept out of click tracking in a measured email; in an email that is not measured, no link is tracked.
A custom tracking domain does not remove the need for consent. It changes the hostname in the tracked link, but the link is still a tracker, and article 82 still applies.
Measurement lives in the dashboard only. There is no export, and there is no statistics method in the SDKs.
Measure outcomes, not only opens
The CNIL's recommendation does not ban tracking pixels. It spells out how article 82 applies to them: outside two narrow exemptions, a person's opens are measured only with their consent. For a SaaS that sends product and transactional emails to people in France, the open rate is no longer a number collected without asking.
For teams that built their email strategy on open rates, this is a shift. For teams that already judged campaigns on conversions, on business events and on delivery, it is a smaller one. The campaign funnel still works and the objective still counts. What changes is that the open rate, in the middle of the funnel, can read "Not measured", or count only part of the recipients.
Notifizz lets each campaign count a business event as its objective, and conditions email measurement on each recipient's consent and preferences. The measurement consent documentation shows how to set it up. Applying measurement consent is one of the components a notification system has to keep working over time, as our guide to building or buying your notification system shows.
Sources
- CNIL, deliberation no. 2026-042 of 12 March 2026 adopting a recommendation on tracking pixels in emails, published in the Journal officiel on 14 April 2026 (Légifrance, in French).
- CNIL, recommendation on tracking pixels in emails (PDF, English version).
- CNIL, Questions-réponses : recommandation relative aux pixels dans les courriers électroniques, 22 July 2026 (in French).
- CNIL, Pixels de suivi dans les courriers électroniques : la CNIL publie ses recommandations pour mieux protéger la vie privée (in French).
- Article 82 of Act no. 78-17 of 6 January 1978 on information technology, data files and civil liberties (Légifrance, in French).
- European Data Protection Board, Guidelines 2/2023 on Technical Scope of Art. 5(3) of ePrivacy Directive.
- Directive 2002/58/EC on privacy and electronic communications (EUR-Lex).
- Regulation (EU) 2016/679, the General Data Protection Regulation (EUR-Lex).
- Notifizz, Open and click measurement: recipient consent (documentation).

