At what segment count does AEP performance visibly degrade and what's governance model to follow to stay ahead of it? | Community
Skip to main content
K
kalarav.vasavada
Level 1
May 4, 2026
Question

At what segment count does AEP performance visibly degrade and what's governance model to follow to stay ahead of it?

  • May 4, 2026
  • 1 reply
  • 20 views

We are about 18 months into a mature AEP + RTCDP implementation and have crossed 400 active segments across business units. We are starting to see non-trivial evaluation latency on streaming segments during peak ingestion windows and our architecture review flagged segment sprawl as a likely contributor.

Adobe's documentation sets technical limits but doesn't offer much guidance on the operational governance model to prevent getting there particularly in a large enterprise where multiple teams (CRM, digital, paid media, product) all have write access to the segment builder.

A few things I'm genuinely curious how others have solved:

  1. Do you enforce a segment approval workflow before publish to production, and if so, is that a process control or is there a technical gate you've built?  
  2. How do you handle segment ownership when the original creator has left the team?
  3. Orphaned segments with no owner are our single biggest hygiene problem.
  4. Have you seen a material performance difference between streaming segments that use only profile attributes vs. those that join experience events?
  5. I suspect event-join segments are our heaviest compute consumers but haven't isolated it cleanly yet.

Curious if anyone has built a segment lifecycle management framework even a lightweight one that scales beyond a single team.

This feels like a gap between what the platform enables and what enterprise operations actually needs.

1 reply

G
GaneshKalamadug
Level 2
May 8, 2026

Very good question. we don’t follow this. We do have many segments and all the scenarios you mentioned. This was my research and we do have a plan in place.  Not sure if this helps you.

  • Yes, enforce a simple approval before production—usually process-driven + restrict publish rights to a smaller team.
  • Make owner (person, not team) mandatory for every segment.
  • Run a weekly check to flag orphaned or unused segments and clean them up regularly.
  • Event-based segments (especially with time windows) are much heavier than profile-only ones.
  • Try to convert event logic into derived profile attributes where possible.
  • Classify segments by complexity and limit heavy ones.
  • Maintain a basic segment registry (owner, use, last used).
  • Add lifecycle rules (e.g., disable if unused for 120–180 days).
  • Key idea: treat segments like managed production assets, not ad hoc builds.
Terms & ConditionsAccessibility statement

Loading footer...