Definitive guide

Why Intent Score Decays: The Math Behind Lead Score Expiration

Tom Regan·Founder & GTM Strategist, Artemis GTM·8 min read·Updated Jul 22, 2026

Quick Answer

Intent scores should decay multiplicatively, not linearly. A common, workable approach multiplies the Intent Score by 0.9 for every 7 days of inactivity (illustrative), zeroes it out near 90 days, and re-evaluates tier. Pause decay during an active opportunity. Linear decay (subtract a fixed number of points per week) tends to over-count old intent and under-count recent intent. Fit does not decay, only intent does. The fastest way to wire this into your CRM is the Lead Scoring agent, built with you inside Claude.

Most lead scoring models track when intent grows. Few track when it expires. The teams that get scoring right run a decay workflow that strips intent points off stale leads automatically, because a 90-day-old "Hot" lead that has not re-engaged is, in most cases, no longer a buyer. This guide covers the decay schedule, why a multiplicative curve beats a linear one, when to pause decay, and how to handle re-engagement. It pairs with the lead scoring pillar guide.

The decay schedule, week by week

The table below is illustrative model math, not measured outcomes. It shows how a multiplicative curve (multiply by 0.9 for every 7 days of inactivity) strips intent from a lead that takes no further action. The "routing implication" column assumes a common Hot threshold, so treat it as an example, not a rule for your data.

Age (days since last action)	Multiplier	Intent retained (illustrative)	Routing implication (example)
0 days (today)	1.00	100%	Full score, route on threshold
7 days	0.90	90%	Still hot if baseline was high
14 days	0.81	81%	Likely still routed on an 80+ baseline
30 days	0.66	66%	Borderline, many drop below threshold
60 days	0.53	53%	Cold tier for most, re-engagement needed
90 days	0.47	47%	Auto-zero in most workflows

As an illustrative worked example, a lead that scored 45 on intent the day they downloaded a case study, with no other action since, moves like this over time (example math, not a measured result):

Day 0: 45 (Hot if Fit is at least 35)
Day 7: 41 (still Hot if Fit is at least 39)
Day 14: 36 (Warm)
Day 30: 30 (Warm or Cold)
Day 60: 24 (Cold)
Day 90: 0 (auto-zeroed, re-evaluate tier)

Why multiplicative beats linear

Linear decay (subtract a fixed amount per week, for example minus 3 points per week) treats the value of old intent and recent intent as constant. That tends to be wrong. Widely cited lead-response work, including Harvard Business Review's 2011 research and MarketingSherpa's lead-aging studies, points to the same directional shape: conversion probability falls steeply in the first 30 days, then flattens. A multiplicative curve matches that shape better than a straight line. These are directional references, not a controlled study of your pipeline.

Concretely, and again as illustrative math: with linear decay of minus 3 per week, a 45-point lead sits at 33 after four weeks (still Hot). With multiplicative decay, the same lead sits at 30 after four weeks (Warm). The multiplicative version tracks the commonly cited conversion curve more closely, where the odds at 30 days are a meaningful fraction of day-0 rather than the near-linear decline a subtraction rule implies.

The 0.9 multiplier is a calibrated default, not a universal constant. In our engagements, dropping it toward 0.85 clears phantom Hot leads faster but starts misclassifying genuine returners as cold, while 0.95 keeps too many stale Hot leads alive. A 0.9 factor is a reasonable starting point; tune it against your own re-engagement data. Directional, not a guarantee.

When to pause decay

Decay should not run in every state. Pause or slow it in these cases:

Lead converts to an opportunity. Set Score_Decay_Pause to true. Behavioral signals shift from marketing engagement to sales engagement, so the intent model is no longer tracking the right thing.
Lead is in an active outbound sequence. If a BDR is working the lead, the score should hold steady so the BDR's effort is reflected in the queue. See BDR queue routing from lead scores.
Lead is in a long-cycle vertical (gov, healthcare, enterprise IT). Buying cycles in these verticals commonly run many months to well over a year. Decay them at half-rate (multiplier around 0.95) instead of pausing entirely.

Re-engagement after dormancy

A lead that decays to Cold and then re-engages is often the highest-quality lead in your queue. They are back of their own volition, usually skipping the top of the funnel and going straight to pricing or demo. The decay workflow brought them down; the new high-intent action adds points fresh.

Many teams under-react to re-engagement because the score still looks low. The fix: in your tier-assignment workflow, branch on both the new total and the velocity of the change. A lead moving from 18 to 33 in one event is a stronger signal than a lead steadily at 65 for two weeks.

Tag re-engaged leads explicitly with last_high_intent_action set to something like "re_engagement_demo". Route them to a senior AE if your routing logic supports skill-based assignment.

The visitor identification gap

Decay only works on leads you can score. A dormant Hot lead that returns to your site as anonymous traffic, visits pricing twice, and leaves without converting never re-fires the intent workflow. The decay marches them down to Cold while their actual intent climbed back to Hot. De-anonymizing returning visitors gives the decay model accurate input on the leads that matter most. See the lead scoring pillar guide for how visitor identification feeds the intent model.

Common mistakes

Linear decay. Subtracting a fixed number of points per week feels intuitive but tends to misfit the commonly cited conversion curve. A 0.9 factor every 7 days is a better-shaped default.
Decaying Fit Score. Firmographics do not expire. A 500-employee fintech is still a 500-employee fintech six months later. Only intent decays.
Forgetting to pause decay during active opps. A late-stage deal drops tier because the buyer stopped reading your blog. Painful and avoidable.
No re-engagement velocity bonus. A lead moving from 18 to 33 in one event is a stronger signal than a lead flat at 65 for two weeks. Many workflows treat them the same.

How to wire decay into your CRM

The decay logic ports cleanly into both major CRMs. Use the template that matches your stack:

HubSpot lead-scoring template: a scheduled Workflow re-checks last-activity date and reapplies the multiplicative factor, with a decay-pause branch for open deals.
Salesforce lead-scoring template: a Scheduled Flow handles the same decay math and the active-opportunity pause.
BDR queue routing from lead scores: how the decayed score orders the working queue so reps touch the freshest intent first.

Methodology and limitations

The decay tables and point walkthroughs on this page are illustrative model math shown to explain the shape of a multiplicative curve, not measured outcomes from a controlled study. The 0.9 factor, the 90-day zero-out, and the tier thresholds are reasonable defaults drawn from our engagements and from widely cited lead-response research; they should be tuned against your own conversion data. Any percentage, retention figure, or conversion claim here is directional and not a guarantee, and it varies by lead source, ICP fit, offer, deal size, and sales process.

Frequently asked questions

Why does intent decay but not fit?

Fit is firmographic, industry, employee count, tech stack. Those do not change week to week. Intent is behavioral, it captures what someone is doing right now. A demo request from 90 days ago tells you little about whether the buyer is still in-market today. Often the champion has changed jobs, the budget has been redirected, or they bought a competitor. Decay is what stops stale intent from crowding the routing queue.

What is the right decay curve?

Many teams default to linear decay (subtract X points per week), and it tends to misfit the data: linear over-counts old intent and under-counts recent intent. A multiplicative curve (multiply by 0.9 every 7 days of inactivity) is a better-shaped default. As illustrative model math, that gives roughly 90% retained at 7 days, 81% at 14 days, 66% at 30 days, and 53% at 60 days. Near 90 days the score is below about half of the original, which is commonly where teams see conversion fall off sharply. These are directional figures, not a guarantee.

Should I decay during the buying cycle if there is an active opp?

No. Set Score_Decay_Pause to true the moment a lead converts to an opportunity. Their behavioral pattern shifts from marketing engagement to sales engagement (calls, emails, doc views in proposals), which does not show up in the same intent signals. You do not want a late-stage deal to drop tier just because they stopped reading your blog.

How do I handle a lead that re-engages after a long dormancy?

Treat re-engagement as a new high-intent action. The decay workflow brought their score down, but the new event (for example a pricing page revisit after four months) adds points fresh. If the action is high-intent enough to cross threshold even from a low base, route them. The story matters: a returning visitor who skips the funnel and goes straight to pricing is often more buyer-ready than a first-time visitor.

What is the revenue impact of skipping decay?

In our engagements, AE queues without decay commonly carry a large share of phantom Hot leads, leads that scored above threshold months ago, never re-engaged, and still trigger alerts. Reps learn to ignore the alerts, so the real Hot leads get the same slow response as the dead ones. That erodes speed-to-lead on the leads that matter. Widely cited lead-response research links faster first touch to materially higher qualification, so a queue clogged with stale Hot leads is a directional drag on conversion, not a guaranteed dollar figure.

Run this play in your own stack

Read the guide, then install the engine.

The Artemis AI GTM Engineer prices this leak in dollars before it recommends anything, then builds the fix with you inside your own Claude. See how an agent installs and buys, or start with the free audit that prices all seven leaks.