stagnation

name	stagnation
description	Safety gate. Detects when a research session is circling — repeated searches returning the same sources, re-analyzing the same competitors without new data, or GTM iterations that keep re-arguing the same channel choice — and suggests a pivot. Runs continuously across the pipeline.

Stagnation — Detect Circular Research

Market research can loop: the same 3 competitors get re-searched with slightly different queries, returning the same articles, while the actual gaps in research/FINDINGS.md stay unverified.

Signals to Watch

1. Repeated searches, same results. Two or more search queries in this session returned overlapping source sets with no new tier-1/2 evidence.
2. Re-reading without re-deciding. The same competitor's FINDINGS.md entry gets revisited multiple times without any field changing.
3. GTM channel indecision. execute or roi_evaluation re-argues the same channel choice 3+ times without new cost/conversion data entering the discussion.
4. Scope creep disguised as thoroughness. Adding "just one more competitor" or "just one more market angle" repeatedly instead of moving to decide/execute with what's already gathered.

Process

1. If 2+ of the above signals trigger, stop and name it explicitly to the user: "This looks like circular research on {topic} — the last N searches returned the same sources."
2. Propose a pivot, one of:
   • Move forward with current evidence, explicitly marking the under-researched point as an "unverified assumption" in the relevant deliverable rather than continuing to search.
   • Change the search angle entirely (different query framing, different source type — e.g. switch from press coverage to pricing pages directly).
   • Ask the user for a missing input only they have (e.g. their own historical CAC data) instead of searching further.
3. Record the pivot decision so it doesn't recur silently.

Write to research/STAGNATION-LOG.md (append):

### {ISO-8601}
**Signal**: {which of the 4 triggered}
**Topic circling**: {competitor / channel / market angle}
**Pivot taken**: {proceed with gap flagged / new search angle / asked user}

Anti-Rationalization

• "One more search might find it" said for the 4th time in a row is the stagnation signal itself, not a reason to continue.
• Thoroughness is valuable up to the point where it stops producing new evidence — after that it's avoidance of a decision.

Evidence

The gate is working when at least one pivot has been logged for any engagement that runs long enough to hit 2+ stagnation signals; a short, efficient engagement with zero log entries is also a valid (good) outcome.

Red Flags

• Same competitor name searched 3+ times in one session with near-identical query text.
• A GTM plan section rewritten 3+ times with no new evidence cited between versions.
• User expresses frustration ("we've been going in circles") before the gate catches it — the gate should trigger first.