AI email tracking changes need a metric-break record

Name the measurement population

For every AI-assisted email program, record the product and account, campaign, audience segment, email address population, stated purpose, consent or exemption basis, preference state, send type, tracking setting, effective timestamp, message volume, and owner. Keep open pixels, click wrappers, delivery events, replies, form submissions, bookings, purchases, and CRM outcomes as separate instruments. The same recipient can be observable on one send and deliberately unobservable on another. A dashboard that merges purpose, consent, and instrumentation states without exposure flags creates a false common denominator.

Write the break before reading the trend

Create a metric-break record when tracking is enabled, disabled, scoped differently, or removed from future AI-generated emails. Preserve the old and new definitions, affected sends and recipients, first and last event times, expected missing events, dashboards and models that consume them, backfill limits, and approval. Annotate time series at the effective boundary and prevent an automated optimizer from learning that an untracked segment is disengaged. Do not impute opens or clicks as observed facts. If historical and current populations cannot be made comparable, show parallel series or restart the baseline rather than smoothing the discontinuity away.

Use alternative evidence carefully

Choose measures available under the approved design, such as delivery and bounce state, direct replies, authenticated site behavior with its own permission basis, qualified inquiries, transactions, retention, or a controlled lift test. Define the numerator, denominator, attribution window, identity match, exclusions, latency, and uncertainty for each. A reply may be meaningful without proving incremental revenue; a conversion can occur without an open event; an automated security scan can create an apparent open. Segment by instrumentation state and preserve channel and customer-experience effects so a privacy choice is not penalized by a model trained on richer tracking elsewhere.

Keep the CMO decision separate from legal advice

The operating decision is which measurement design the organization will use, how it will honor its approved purpose and consent strategy, and how decision makers will interpret missing data. Route jurisdiction, purpose, disclosure, consent, withdrawal, and retention questions to qualified owners rather than treating a recommendation as a case-specific legal conclusion. Require marketing operations, analytics, privacy, legal, channel, and campaign owners to approve the metric-break treatment before performance reviews, budget reallocations, audience suppression, or automated optimization rely on it. The CNIL describes its recommendation as non-regulatory and non-exhaustive; it does not determine another jurisdiction's rules or demonstrate campaign effect.

Turn this source into a reviewable decision

For AI for CMOs, use this briefing as a dated decision record rather than a substitute for the source. Preserve Recommendation on tracking pixels in emails, the exact URL, the September 14, 2026 review date, the supported facts above, the editorial interpretation, the limitations, and any buyer-specific evidence. Link that record to the decisions most directly affected: Measurement and performance explanation; Customer journeys and personalization; Media planning and activation; Brand, disclosure, and synthetic-media risk. State whether the source changes the scope, evidence requirement, control, sequence, or only the language used to describe the decision.

Before action, name the accountable owner, affected population and workflow, exact offering or configuration, source data and rights, human decision point, exception and appeal path, complete cost, expected benefit, failure and stop conditions, retained evidence, and next review date. Keep official facts, provider statements, buyer observations, representative tests, measured outcomes, editorial inferences, and unknowns visibly separate. Reopen the record when the source, offer, model, integration, data, policy, population, responsible person, or measured result changes.

Limitations and unknowns

The source is an English courtesy translation of a French authority recommendation, and the French version prevails if the texts differ. The CNIL says the recommendation is non-regulatory and non-exhaustive; it does not establish another jurisdiction's rules, a sender's purpose, valid consent or exemption, configured tracking state, event completeness, dashboard logic, attribution, campaign quality, or business outcome. Current send populations, purpose records, consent and withdrawal evidence, configuration exports, raw delivery and engagement events, dashboard definitions, alternate outcome evidence, and qualified marketing operations, analytics, privacy, legal, security, accessibility, and regional review control.