List Segmentation and Consent Architecture for Email

Most accounts I open carry between 60 and 200 saved segments and use about nine of them. The count is not really the problem. The problem is that no two of those segments define an active subscriber the same way, so the flow report, the campaign report and the board deck disagree and nobody can say which one is wrong. The first thing I deliver is a single written definition of active, and everything else in the account hangs off it. Engagement is measured on clicks, not opens: Apple Mail Privacy Protection pre-fetches images, so an open is now a weak positive and a strong negative. Engagement here means a click, a site session, or a purchase or donation. An open counts only as evidence that the address still exists. Bands run 0 to 30 days, 31 to 90, 91 to 180 and 181 to 365. The 181-to-365 band typically carries five to fifteen times the complaint rate of the 0-to-90 band while contributing a fraction of the revenue, and removing it from the send is what raises placement for the people who pay. The sunset ladder: Past 365 days, a re-permission sequence with a real choice in it, then suppression. Suppression is not deletion. The profile and its consent record stay for the retention period you have declared; sending stops. Accounts that skip this step are the ones where a Gmail placement problem appears eight months later with no obvious cause. Consent as a segment dimension, not a checkbox: Spanish commercial email runs on prior consent under Article 21 of the LSSI-CE, with a separate basis for an existing customer relationship offering similar products. Those two bases are different audiences with different rights and they cannot sit in one list. Every profile carries its basis, timestamp, capture source and the wording it was shown. The preference centre offers topic and frequency, and the frequency option earns its place: giving people a way to receive less removes more unsubscribes than any subject-line work I have done. By sector this looks different: In fintech, product eligibility is a segmentation problem before it is a targeting one. Where a product cannot be promoted to an ineligible profile, that rule belongs inside the segment definition, where it is enforced, rather than in a marketer's memory at four o'clock on a send day. In nonprofit work, donors are not customers. Recency, frequency and amount replace purchase behaviour. An annual giver looks lapsed at 13 months and is not. Mid-level donors should be structurally excluded from the mass appeal that would insult them, and recurring donors should never receive an acquisition ask. In beauty, the useful segment is the replenishment window derived per SKU from unit size and use rate, joined to a shade or skin-concern attribute collected at signup. That attribute is given by the customer and it is reliable, which is more than can be said for an inferred one. What I refuse: Data appends and third-party enrichment of email lists. They corrupt the consent record, and no Spanish regulator treats a purchased interest signal as consent. Look-alike exports built from a file whose consent cannot be evidenced. Personas as segments, because "the busy professional" is not a filter and cannot be written as one. And predictive lifetime value used as a segment boundary in an account with under a year of purchase history, where the model is fitting noise and will be defended anyway because it has a number in it. Boundaries: This engagement does not build flows or run campaigns, though it defines the audiences both use. It does not implement a CDP. It does not include warehouse modelling beyond specifying what a reverse ETL job has to push into the platform, and how often. Not for an account under roughly 5,000 engaged profiles. At that size the segments that matter fit on one screen and you do not need me to draw them.