Sampling, Selection, Summarization: Scaling Feedback Without Losing Signal
TL;DR: When feedback volume grows, you need rules for what enters the system, what gets sampled, and what gets summarized—without hiding reality.
The scaling problem
At 50 inputs/week, you can read everything.
At 500/week, you can’t. If you don’t change the process, you’ll either drown or over-summarize.
Sampling rules
Use sampling to keep representation:
- sample by segment (SMB, mid-market, enterprise)
- sample by journey stage (onboarding, daily use, renewal)
- sample by channel (support, sales, research)
Selection rules
Select based on:
- severity (blocks task vs annoyance)
- frequency (seen across sources)
- leverage (fix improves multiple workflows)
Summarization rules
Summarize on cadence:
- weekly: candidate themes
- monthly: top themes + evidence pack
- quarterly: strategic narrative
Takeaways
Scale requires rules. Rules create trust in the signal.