listen · lead scoring and merging

Tiers and contact policy

Hot, Warm, Watch and who never gets contacted.

In Lead scoring and merging

stage 7 · lead scoring and merging · tiers and contact policy

Hot, Warm and Watch decide when a lead gets a move. The contact policy decides whether it can get one at all. Both live on the person, so they follow every handle after merging.

signal → lead → tiers and policy → playbook

Tiers are service levels — Hot means today, Warm means this week, Watch means keep listening. Nothing else.

Policy beats tier — A Hot lead in cooling-off gets nothing. The check runs before every move, on every channel.

Never-contact is forever — Set by a stop request, a lead rule or the founder. Nothing automatic ever clears it.

what it is, in industry terms

Three tiers decide when; the contact policy decides whether.

Tiers turn a score into a service level a solo founder can keep. The contact policy is the guard every move passes first. Together they're why a distribution agent can act fast on Hot leads without ever spamming anyone.

Lead tier — also: lead grade, priority bucket · A small number of buckets that decide the service level: who gets a move today, this week, or not yet.

Speed to lead — response-time SLA · How fast a tier must be acted on. A Hot lead's buying window is measured in hours.

Contact policy — frequency caps, cadence rules · Limits that apply before any message: how often, how close together, which channels, which hours.

Suppression list — do-not-contact, never-contact · People who must never get a move. Set by the founder, a stop request or a lead rule. Never cleared automatically.

Cooling-off — rest period · A window after a touch with no reply when no new touch is allowed.

Override — manual tier · The founder sets a tier by hand. Kept as a label: it says where the rule was wrong.

how it works

From a lead to today's list, through two checks.

Check the person
never-contact or a lead rule (customer, never, competitor): out
Apply the public-pain rule
all three fit lines hold: Hot or Warm. Any fails or is unknown: Watch
Split Hot from Warm
proposed: fresh strong or stacked signals are Hot; older, weaker or single are Warm
Record the change
every tier change with its cause and rule version
Check the contact policy
caps, cooling-off, quiet hours, channel allowed, at the person level
Release to the list
Hot today, Warm this week, Watch keeps listening

Check the person — never-contact or a lead rule (customer, never, competitor): out
Apply the public-pain rule — all three fit lines hold: Hot or Warm. Any fails or is unknown: Watch
Split Hot from Warm — proposed: fresh strong or stacked signals are Hot; older, weaker or single are Warm
Record the change — every tier change with its cause and rule version
Check the contact policy — caps, cooling-off, quiet hours, channel allowed, at the person level
Release to the list — Hot today, Warm this week, Watch keeps listening

Tier	Service level	Rule
Hot	a move today	all three public-pain fit lines hold, and the signal is fresh and strong, or stacked
Warm	a move this week	all three fit lines hold, but the signal is older, weaker or single
Watch	no move yet	any fit line fails or is unknown; a second signal or a confirmed line can promote

the objects

A tier change, and the policy state of a person.

tier_change

Field	Type	What it holds
change_id	uuid	One tier change.
lead_id	uuid	Whose tier changed.
from, to	hot \| warm \| watch \| out	The move. `out` covers never-contact and lead rules.
cause	enum	new signal, signal aged, fit line confirmed, fit line failed, founder override, lead rule, stop request.
rule_version	string	The tier rule that made the call, e.g. `public-pain.2026-10-09`.
by	system \| founder	Who changed it.
at	timestamp	When.

contact_policy (per person)

Field	Type	What it holds
person_id	uuid	Policy lives on the person, so it follows every handle after merging.
last_touch_at	timestamp	The most recent move, any channel.
touches_7d	int	Moves in the last 7 days, all channels.
cooling_until	timestamp	No move before this.
channels_allowed	list	Public reply only, DM allowed (after they engaged), email (only if they gave it).
quiet_hours	local window	No sends at night in their timezone, when it's known.
never_contact	bool + reason + source	Founder, stop request or lead rule. Sticky across re-imports.

sample, as data

{
 "tier_change": {
 "change_id": "tc_301",
 "lead_id": "ld_51aa",
 "from": "watch",
 "to": "warm",
 "cause": "fit line confirmed",
 "rule_version": "public-pain.2026-10-09",
 "by": "system",
 "at": "2026-10-17T13:02Z"
 },
 "contact_policy": {
 "person_id": "pn_2c7d",
 "last_touch_at": null,
 "touches_7d": 0,
 "cooling_until": null,
 "channels_allowed": [
 "public_reply"
 ],
 "quiet_hours": "22:00-08:00 local",
 "never_contact": {
 "value": false
 }
 }
}

the catalog

Sixteen tier changes and policy rules.

tier.new-strong — tier change

Fresh strong signal

move: watch → hot or warm
when: a strong signal under its half-life arrives and all fit lines hold
note: Hot if fresh and strong; Warm otherwise

tier.stacked — tier change

Second signal stacks

move: warm → hot
when: a signal from a different detector or trigger arrives inside the window
note: stacking is the strongest promotion

tier.aged — tier change

Signal aged past its window

move: hot → warm → watch
when: the strongest signal passes one, then two half-lives with no new one
note: keeps yesterday's Hot from crowding today's list

tier.line-confirmed — tier change

Fit line confirmed

move: watch → warm
when: enrichment or the founder confirms the unknown fit line
note: the commonest promotion out of Watch

tier.line-failed — tier change

Fit line failed

move: any → watch
when: they turn out not to sell to people, or the product isn't live
note: demote, keep listening

tier.override — tier change

Founder override

move: any → any
when: the founder sets the tier by hand
note: kept as a label: the rule was wrong here

tier.replied — tier change

They replied

move: any → in conversation
when: a reply to a move arrives
note: leaves the tier list; the conversation's own next step takes over

tier.joined — tier change

Joined or paid

move: any → converted
when: a waitlist join or a payment matches the person
note: closes the lead; it becomes an outcome label

policy.stop — policy rule

Stop request

move: any → out, sticky
when: they say stop, block or report
note: never cleared automatically; logged as negative feedback

policy.lead-rule — policy rule

Lead rule match

move: any → out
when: the founder's rules say customer, never a customer, or competitor
note: never becomes a lead

policy.cap-7d — policy rule

Weekly cap

move: blocks a move
when: proposed: a third touch inside 7 days, across all channels
note: caps apply to the person, after merging

policy.cooling — policy rule

Cooling-off

move: blocks a move
when: proposed: a touch with no reply in the last 3 days
note: one unanswered message is a fact; two is a pattern

policy.dm-gate — policy rule

DM gate

move: blocks a channel
when: they haven't engaged with you publicly yet
note: public reply first; a DM only after they answer

policy.quiet — policy rule

Quiet hours

move: delays a move
when: it's night where they are, when their timezone is known
note: drafts wait for morning

policy.two-in-a-row — policy rule

No reply twice

move: watch, cooling 30 days
when: proposed: two moves, no reply
note: back to listening; a fresh signal can restart

policy.channel-limit — policy rule

Platform limits

move: blocks a channel
when: the platform's own reply or DM limits are near
note: never let a cap bite in the middle of a sequence

illustrative example

One morning, before launch.

With one founder spending about 20 minutes a day, tiers have to keep Hot short. A rule of thumb to start with: if Hot holds more than ten people on a given morning, the Hot cut is too loose, and the fix is to require a stacked or very fresh signal, not to work faster.

ONE MORNING · EXAMPLE

Hot: 4 people: 2 asked for a tool in the last 12 hours, 1 stacked (Reddit + X reply), 1 commented the keyword
Warm: 23 people, a public reply each this week
Watch: the rest; any fresh signal can promote them
blocked: 2 by cooling-off, 1 DM gate, 1 never-contact (said stop last week)
result: 4 moves today, each logged with tier, rank and the policy state at send time

failure modes

Six ways tiering and policy go wrong.

Everyone is Hot — Loose rules put fifty people in Hot and the list stops meaning anything. · catch ita daily Hot count with an alert; tighten the rule when it grows

Stale Hot — A lead stays Hot for a week because nothing demotes it. · catch itage-based demotion runs every day, not only on new signals

Policy bypass by channel — Capped on X, the same person gets a Reddit reply because the two handles weren't merged. · catch itcaps live on the person; unmerged duplicates are counted as incidents

Never-contact lost — A re-import or a new handle brings back someone who said stop. · catch itnever-contact sits on the person and is checked against every new claim

Overrides that teach nothing — The founder fixes tiers by hand and the rule never learns why. · catch itstore every override with the rule's tier beside it; review them monthly

Quiet hours guessed wrong — A timezone guessed from a single post sends at 3am. · catch itonly apply quiet hours when the timezone is known; otherwise default to your daytime

data to capture

Every change and every block, with its reason.

record

every tier change with cause, rule version and who made it
the contact policy state at the moment of every move, and every block with the rule that blocked it
founder overrides, with the rule's tier beside the chosen one
stop requests and negative feedback, tied to the touch that caused them

label

override: the rule's tier was wrong for this person
replied, joined or paid by tier: does Hot really outperform Warm?
stop requests per 100 touches by tier and channel (a guardrail)

trains

calibrated tier cutoffs, so each tier has a measured reply rate
a cap and cooling-off setting learned from stop requests against replies

test

offline: replay the tier rule on past leads and compare the reply rate per tier
online: test one policy knob at a time (cooling 3 days vs 5) with leads randomized, stop requests as the guardrail

Handles, products and scores in examples are invented. Merge only on public or first-party data, keep never-contact sticky, and a person always sends the first message.