Segment Sunset Reviews: Retiring Segments That Stopped Earning Their Keep
Ask a marketing team to list its segments and you will get a confident answer about creation: who built each one, when, and why. Ask which segments were retired last year and the room goes quiet. Segments are born in workshops and die almost nowhere, so the system fills with audiences nobody targets, rules nobody maintains, and definitions that quietly stopped matching the customers they name.
Every segment carries a maintenance cost
A segment is not a label. It is a rule that must stay true while products, prices, tracking, and the customer base all move underneath it. A segment built on a product category that was renamed, a behaviour the app no longer logs, or a campaign code retired two promotions ago does not fail loudly. It fails by drifting, still countable, still selectable in the tool, describing a group that no longer exists in the form its campaigns assume.
The sunset review prices that drift. For each segment, ask three questions on evidence: when was it last used in a live campaign, what treatment does it receive that no other segment receives, and what did that treatment return the last two times it ran. A segment that cannot answer the second question is a report wearing a segment's clothes, and it should be demoted to one before it wastes a campaign slot.
The review meeting, run properly
Bring the data, not the owners' defences. Print every active segment with its size, last use, distinct treatment, and last measured result, and sort by last use. The bottom of that list is the agenda. Segments unused for two quarters need a positive case to survive, made in the meeting, not a presumption of innocence because someone senior built them in a strategy offsite.
Retirement needs a ceremony smaller than creation and just as official: the segment is archived with its definition and its history, its owner records where its members now fall, and any automated campaign referencing it is rewired the same week. Deleting a segment that a live automation still points at is how a tidy up becomes an incident, so the dependency check is not bureaucracy. It is the whole job.
What usually survives, and what should not
Behavioural segments tied to current products and a distinct treatment tend to survive every review, because their answers are easy. The casualties are predictable: event segments built for a launch two years past, demographic segments superseded by behavioural ones that cover the same people better, and the duplicate pair, two segments with different names and ninety percent shared membership, each with its own owner, each mailing the same customers.
Duplicates deserve special firmness. Merge them under one name and one owner in the meeting, not as a follow up task. Follow up tasks are where merges go to be postponed until the next review counts the same duplication again, with a year's extra mail on the same customers as the fee for the delay.
Make creation pay the sunset toll
The lasting fix is at the door. New segments should be proposed with an owner, a distinct treatment, and a review date, recorded in the same register the sunset review uses. That one habit turns the segment list from an attic into an inventory: everything in it is owned, purposeful, and scheduled to justify itself again. Teams resist the paperwork for about one cycle, until the first review where a zombie segment they inherited gets retired by evidence instead of by their own exhausting archaeology.
A shorter segment list is not a smaller ambition. It is the same ambition with the dead weight removed, and every campaign built afterwards starts from definitions somebody has recently sworn are true.
Keep reading
Related guides on this site: Test and Control Segments: Holding Out a Clean Comparison Group Clearance Segments: Finding Buyers for End-of-Season Stock Without Training Discount Waiting Practical answers about personalization.
Frequently asked questions
What is a segment sunset review?
A periodic review, usually quarterly or twice a year, that scores every active segment on last use, distinct treatment, and measured return, and formally retires segments that no longer justify their maintenance cost.
Why do unused segments cause harm if nobody mails them?
Because their rules drift as products, tracking, and prices change, while they remain selectable in the tool. Sooner or later someone builds a campaign on a definition that quietly stopped describing the customers it names.
How should a segment be retired safely?
Archive its definition and history, record where its members now fall, and check and rewire every live automation or campaign that references it in the same week. The dependency check prevents a tidy up from breaking a running programme.
How can segment sprawl be prevented?
Require every new segment to arrive with an owner, a distinct treatment, and a review date in a shared register. Segments created as inventory get reviewed as inventory, instead of accumulating as attic clutter.