High-intent web behavior should decay quickly unless the buyer repeats it. One pricing visit is a spark, not a six-month mandate to chase.
// GTM CONTROL PLANES · TRUST DENSITY · June 5, 2026
GTM Control Planes Need Signal Half-Lives Before Scoring Turns Stale
A buying signal is not a permanent fact. If your agents cannot tell fresh evidence from stale evidence, automation turns yesterday's intent into today's bad judgment.
Scoring needs an expiration date.
Most GTM systems treat signals as additive. A webinar visit, funding event, job change, product-page spike, support ticket, and enrichment update all become points in a score. The score rises, the agent routes the account, and everyone assumes the machine is reacting to current reality.
That assumption breaks fast. A CFO researching pricing in January may be irrelevant in June. A hiring signal may matter for thirty days. A competitor mention may matter for a week. A board-triggered budget window may close before sales even sees the account.
A signal half-life defines how quickly each piece of evidence loses authority. It does not delete history. It prevents old history from pretending to be live buying intent.
Stale truth is still dangerous.
Agents do not need hallucinations to create bad GTM motion. They only need clean data with no age penalty. Old evidence can trigger fresh outreach, wrong routing, inflated forecasts, and exhausted sales capacity.
Decay creates trust density.
When every signal has a clock, reps can trust that machine recommendations are built from current evidence instead of accumulated noise. Fewer actions become more credible.
Firmographic fit decays slowly, but not never. Headcount, tech stack, region, and ownership changes should refresh before they drive routing.
Budget, hiring, compliance, and renewal windows are perishable. The control plane should know when a timing signal is past its useful life.
Human context carries weight, but it also ages. A warm intro, champion note, or past meeting should expire into review before an agent acts on it.
Grounded examples
Where stale scoring burns capacity
The zombie MQL
An account hits the threshold because three old campaign touches never decayed. Sales gets a high-priority route with no current buying evidence.
The outdated champion
An agent drafts outreach around a champion who changed jobs months ago. The rep looks careless because the system preserved context without freshness.
The false segment
A nurture worker keeps treating a company as expansion-ready because an enrichment field stayed static after the team downsized and froze spend.
What belongs in a signal half-life map
The map should be simple enough for operators to maintain and explicit enough for agents to enforce. Every signal class gets authority, decay, refresh, and action rules.
Signal class
The evidence type: intent, fit, timing, relationship, risk, product usage, enrichment, or human note.
Separate them because each ages differently. A pricing spike and an employee count field should not share the same clock.
Decay rule
The period when authority drops: immediate, seven days, thirty days, one quarter, or until refreshed by a verified event.
Decay should lower confidence, suppress action, require review, or change the agent from execution mode to research mode.
Refresh source
The evidence allowed to renew the signal: repeated behavior, source-system update, human confirmation, call outcome, or verified enrichment.
Do not let the same old event refresh itself. Renewal needs new evidence.
Action limit
What the agent may do while the signal is fresh, aging, stale, or contradicted.
Fresh may route. Aging may research. Stale may archive. Contradicted should stop the workflow and explain why.
Operator tests
Four checks for this week
Pick ten agent-routed accounts and identify the newest signal that justified action. If the newest signal is old, the score is lying.
Show signal age beside the score in the workspace where humans approve, reject, or inherit agent decisions.
For one workflow, decide exactly what the agent does when its strongest signal expires: pause, research, downgrade, or ask for review.
Track how many touches, routes, and forecasts were prevented because the evidence aged out. Restraint is a control-plane metric.
The question is not whether the signal was ever true. The question is whether it is still allowed to move work.
Concrete takeaway for this week
Add half-lives to the three signal classes that trigger the most GTM automation.
- Classify the signal. Separate intent, fit, timing, relationship, and enrichment evidence.
- Attach a clock. Decide how long each signal can drive action without renewal.
- Set stale-state behavior. Tell the agent to pause, research, downgrade, or request review when evidence ages out.
- Make freshness visible. Show signal age wherever humans see scores, routes, and recommendations.
GTM control planes do not become trustworthy by collecting more signals. They become trustworthy when they know which signals still deserve authority.