understand · icp and the trigger window

The trigger window

How long a buying window stays open.

In ICP and the trigger window

wiki · stage 3 · icp and trigger detection · component 2 of 2

Fit says who could buy. The window says when. For a founder who just launched, the window opens about three days after launch and closes near day 45, peaking in days 5 to 21 with a second spike at 28 to 35. Before it, a message is noise; after it, it reads as a rebuke.

Timing is a separate layer.

Fit changes slowly and lives on the lead. Timing goes stale in days and lives on the signal.

Every trigger has a clock.

Each trigger type gets a freshness window, fitted from data once there's enough, guessed and labelled until then.

The exit is not a target.

"Should I give up" is real pain, but it marks the window closing. Help in public; don't sell.

what it is

The window of dissatisfaction, in industry terms

Industry names: trigger event, buying window, timing in signal-based selling. Craig Elias's term is the Window of Dissatisfaction: "unhappy with Status Quo but not searching for Alternatives yet." It opens when something breaks the person's belief that their current approach works, and closes when they fix it, buy something or give up.

For the launch trigger the window has a measured shape from the Sep 19 research: 26 dated founder posts across five weeks. Directional, not a tracked cohort, so the bars on this page are priorities, not rates.

Window opens — forcing event · something breaks: traffic dies on day 2 to 4, trials expire around day 28 to 35.

Peak — highest priority · days 5 to 21, tactical questions: "where did your first users come from?"

Second spike — the money doesn't arrive · days 28 to 35, first-month numbers.

Exit — window closes · day 60 and later; existential posts.

Freshness — fresh_for · how long a signal stays actionable; a field on every detector.

Decay curve — reply rate by age · the data that replaces a guessed window with a fitted one.

how it works

How the window becomes a ranking

The launch date is the trigger, and it comes free from launch directories with no inference. Pain phrases since launch confirm the window is open. Together they rank the queue: who is likely to reply this week.

Find the launch date — launch feeds: Show HN, PeerPush, Product Hunt
Compute days since — day 0 = launch; place the founder on the curve
Check for pain — a pain phrase posted since launching?
Score the window — peak, secondary, cooling or closed
Add fit signals — paid slot, TrustMRR band, follower count
Gate on reachability — founder handle found, or skip
Rank the queue — likely to reply this week first
Expire — drop from the queue when the window closes

Find the launch date — launch feeds: Show HN, PeerPush, Product Hunt
Compute days since — day 0 = launch; place the founder on the curve
Check for pain — a pain phrase posted since launching?
Score the window — peak, secondary, cooling or closed
Add fit signals — paid slot, TrustMRR band, follower count
Gate on reachability — founder handle found, or skip
Rank the queue — likely to reply this week first
Expire — drop from the queue when the window closes

The curve

day 0–2 · Euphoric — send nothing
launch day traffic; the founder is celebrating

day 3–10 · "Traffic died, what now" — primary
the traffic dies, day 2 to 4; tactical questions start

day 11–21 · Trial purgatory — activation help
signups don't activate; the founder refreshes analytics

day 28–35 · "My first month numbers" — secondary
trials expire; the money doesn't arrive

day 36–60 · Cooling — light touch
the founder is deciding what to do next

day 60+ · Rationalizing an exit — nothing
existential posts: "should I give up?"

Two forcing events shape it. The traffic dies on day 2 to 4: day zero is roughly 60% of a launch's traffic, days 1 to 7 about 25%. The money doesn't arrive on day 28 to 35: about 35 to 45% of signups activate within 7 days and 5 to 10% of those pay within 30, so a 50-signup launch produces one or two customers, often zero.

the object

Timing on a signal

Timing is stored on the signal, not the lead, because it goes stale on its own clock. The lead reads it when it is ranked.

signal_timing

field	type	what it holds
`trigger_id`	enum	which trigger, e.g. launched_recently
`event_at`	timestamp	when the trigger happened: the launch date or the post time
`captured_at`	timestamp	when your system saw it; the gap is detection lag
`age_days`	derived	days since event_at, at ranking time
`phase`	enum	pre, peak, trial, second_spike, cooling, closed
`fresh_for_days`	int	the window for this trigger; guessed until fitted
`expires_at`	timestamp	event_at + fresh_for
`window_source`	enum	guess, research, fitted

sample

{
 "trigger_id": "launched_recently",
 "event_at": "2026-10-03T16:00:00Z",
 "captured_at": "2026-10-03T18:20:00Z",
 "age_days": 13,
 "phase": "trial",
 "fresh_for_days": 45,
 "expires_at": "2026-11-17T16:00:00Z",
 "window_source": "research"
}

ranking signals

What ranks the queue inside the window

The Sep 19 research's weights. 'Gate' means required: without a founder handle there is no one to reach.

timing

3–21 days since launch

weight: highest
why: the window at its peak

budget

Paid for the launch slot

weight: high
why: proven budget and dissatisfaction

pain

Posted a pain phrase since launching

weight: high
why: stated, not inferred

pain

Low upvotes on a paid tier

weight: medium
why: paid and it didn't work: the sharpest state

fit

TrustMRR row at $0–500

weight: medium
why: confirms the segment

fit

Under 2,000 followers

weight: medium
why: no distribution of their own

timing

28–35 days since launch

weight: medium
why: the second spike

gate

Founder handle found

weight: gate
why: required to reach them

Phrases that confirm the window is open

Ranked by frequency times clarity. The compound that matters most: an elapsed-time phrase and a zero in the same post.

"launched N days ago" + "0 users" — the whole ballgame: elapsed time and a zero together

"crickets" / "went quiet" — on a first-person launch

"nobody knows it exists" — the distribution gap, named

"I'm a builder, not a marketer" — reliable on Reddit, not on X

"refreshing analytics" — the purest tell

"should I give up" — real pain, but the window's exit: help, don't sell

failure modes

How it goes wrong, and how to catch it

Messaging the euphoric — Day 1 congratulations followed by a pitch. · Catch it: nothing before day 3; the phase field blocks it.

Selling to the exit — "Should I give up" treated as a hot lead. · Catch it: phase closed routes to help-in-public, never an offer.

Guessed windows treated as facts — A 45-day window applied to every trigger type. · Catch it: window_source shows guess vs fitted; fit it once replies exist.

Detection lag — A signal caught on day 9 is ranked as if caught on day 2. · Catch it: store event_at and captured_at; rank on age, not capture time.

Retrospective posts — A war story from last year read as current pain. · Catch it: past-perfect tense and no current numbers: discard as retrospective.

data to capture

Record it so a model can learn from it later

Every decision on this page is training data if it is recorded in the guide's shape from day one.

1 · log the decision · 2 · version whatever decided · 3 · freeze the inputs · 4 · keep the negatives · 5 · log the odds · 6 · every human touch is a label · 7 · one chain of ids · 8 · keep the raw

record

event_at and captured_at on every signal (posted time and capture time)
the phase and age at the moment the lead was ranked, frozen
reply and signup outcomes per signal, with age at touch
discards for being outside the window

label

reply and signup per trigger, by age in days

trains

a fitted freshness window per trigger (the decay curve) to replace the guessed one, and a value ranking of triggers

test

knob: each trigger's freshness window (for example, 24 hours vs 72 hours for a pain post)
mode: offline from the decay curve first, then by lead cohort
win rule: replies per touch, holding playbook and copy constant