Segment decay: when audiences go stale and quietly stop performing
Every segment is a snapshot of a moving population. People who were high-intent in spring are out of the market by autumn, new buyers qualify who were never added, and rules written for last year catalog describe this year shoppers badly. Segment decay is normal. The failure is having no process that notices it.
What decay looks like in the numbers
The early sign is engagement drift inside a stable segment: open and click rates sliding a point at a time while list size stays flat. Because membership looks constant, teams blame creative, when the audience itself has quietly rotated into people the content no longer fits.
Conversion decay lags engagement decay. By the time revenue per send from a segment drops clearly, the segment has usually been mailing cooled intent for months, and the unsubscribes it collected on the way are a permanent cost.
The three kinds of staleness
Membership staleness is people who no longer meet the spirit of the rule: the lapsed segment still holding last year buyers because the window was set to twenty-four months in a twelve-month category.
Intent staleness is subtler: members still match the rule but their reason has passed, like gift buyers in January segments built for December. Data staleness is the third kind: the attributes feeding the rule stopped updating, so the segment describes a feed failure as if it were a shopper.
Refresh cadences that match reality
High-intent segments need short windows and frequent recomputation, ideally continuous. Evergreen segments like category buyers can refresh weekly or monthly, provided the qualifying window reflects the real repurchase cycle.
Write the expiry assumption into the segment definition itself: the rule, the window, the refresh schedule, and the owner. A segment nobody owns decays at the speed of the business changing around it.
Auditing a segment library
Twice a year, list every active segment with its size trend, last rule change, and performance trend. Segments that are shrinking, untouched for a year, or overlapping another segment by more than eighty percent are audit findings, not assets.
Sampling members is the ground truth check. Pull twenty profiles from a segment and ask whether a human would say they belong. Rule logic that passes its own tests can still produce a membership a marketer would not recognize.
Rebuilding versus retiring
Rebuild when the underlying audience still exists and the rule is the problem: tighten the window, add the missing exclusion, fix the stale feed. Retire when the campaign idea itself has passed, however elegant the segment was.
Keep retired definitions archived with their performance history. Next year team will otherwise rebuild the same segment from scratch and rediscover the same decay curve at full price.
Frequently asked questions
How often should segments be refreshed?
Intent-based segments should recompute continuously or daily, while stable buyer segments can refresh weekly or monthly. The qualifying window matters more than the schedule: it should match the real buying cycle.
What is the first sign a segment has decayed?
Slow engagement drift with stable membership. When opens and clicks slide gradually while the list size does not change, suspect the audience composition before blaming creative.
Should old segments be deleted or archived?
Archive the definition and its performance history, but remove it from active use. Keeping decayed segments sendable invites accidental mailings to audiences nobody has checked.
Can segments overlap safely?
Yes, with clear priority rules for which segment content wins when a shopper qualifies for several. Unmanaged overlap means the same person gets competing messages chosen by send order.