Segment Refresh Cadence: How Stale Can a Segment Get Before It Lies?
Segments feel like facts once they are named in a dashboard. VIPs, at risk, new buyers. Underneath, every segment is a query run at a moment in time against behaviour that keeps moving, and the moment a campaign mails to a segment computed three weeks ago, it is mailing to a description of people who have since bought, churned, returned things, and changed their minds.
Decay rates differ by segment definition
Behavioural segments decay fastest. A browsed but did not buy segment built for a weekend campaign is mostly wrong by next weekend, because the behaviour it captures has a shelf life measured in days. Lifecycle segments, first purchase year, tenure bands, decay slowly and can ride a monthly refresh without real harm.
Value segments sit in between and decay in jumps: one large order or one large return can move a customer across a monetary threshold overnight. If a high value segment gates a benefit like free expedited shipping, a stale refresh pays that benefit to people who returned their way out of the segment weeks ago.
Refresh before the send, not on a calendar alone
The calendar habit, segments refresh nightly, weekly, monthly, is fine as a baseline and wrong as a guarantee. The rule that matters operationally is freshness at the moment of use: a campaign should check the segment's computed date and the segment should recompute, or at minimum revalidate membership, before a high stakes send or an offer with real margin cost.
Suppression lists deserve the same discipline in reverse. A suppression built from last month's unsubscribes and complaints is a compliance problem, not a marketing inefficiency, if it misses this month's opt outs. Legal and deliverability data should refresh on the tightest cadence in the building.
Streaming is not always the answer
Real time segment membership is sold as the cure, and for triggered messages based on current behaviour it often is. For planned campaigns it buys complexity and a new failure mode: membership that changes during the send itself, producing partial audiences nobody can reproduce afterward when finance asks who got the offer.
Batch refresh with a timestamp, a frozen audience file per send, and a documented computed at time is auditable in a way streaming membership is not. For most stores the right answer is boring: daily refresh for campaign segments, frozen snapshots for anything involving an offer, and real time only where a trigger genuinely depends on the last few minutes.
Name the staleness on the segment itself
Dashboards should show each segment's computed date next to its size, and a size that has not moved in a month on a supposedly behavioural segment should be treated as a broken pipeline, not stability.
Teams that can see staleness manage it. Teams that see only a name and a number will mail to a ghost segment eventually, and learn about refresh cadence from the complaint thread.
Keep reading
Related guides on this site: Lookback Windows: The Quiet Setting That Changes Who Is in Every Segment How Small Can a Segment Be Before It Is Just Noise Wearing a Name Practical answers about personalization.
Frequently asked questions
How often should segments refresh?
It depends on decay rate. Behavioural segments can go stale in days, value segments shift with each large order or return, and lifecycle segments tolerate monthly refresh. Campaign audiences should be fresh at send time regardless of the calendar.
Is real time segmentation always better?
No. It suits behaviour triggered messages, but planned campaigns benefit from a frozen, timestamped audience file that can be audited later. Streaming membership that changes mid send is hard to reproduce or explain.
Why does segment staleness matter for compliance?
Suppression lists built from old data can miss recent opt outs and complaints, turning a marketing shortcut into a legal and deliverability problem. Suppressions deserve the fastest refresh in the system.