Score Decay

Score decay (also called recency decay) is the mechanism that reduces a lead’s score over time so that old activity doesn’t keep them artificially hot. Without decay, a pricing-page view from six months ago counts the same as one from yesterday — and your “hot” list silently fills with cold leads.

Score decay is the difference between a scoring model that stays honest as time passes and one that quietly accumulates noise until sales stops trusting the score.

How score decay works

Most lead scoring systems implement decay in one of three ways:

1. Group-level percentage step

The simplest model. Pick a cadence (every 1, 3, 6, or 12 months) and a percentage reduction (e.g. -25%/month). Every event in the group decays linearly at that rate.

This is what HubSpot’s Modern Lead Scoring (Aug 2025) ships. A pricing-page view, a whitepaper download, and a webinar registration all decay at the same rate, even though they’re meaningfully different signals with different intent durations.

2. Time-window cutoff

Activity older than X days is dropped entirely. “Score only counts engagement from the last 90 days.” Crude but effective; a lot of legacy systems work this way.

3. Per-event half-life decay

Each event type has its own decay curve. A demo request might have a 90-day half-life (still scoring 50% after 3 months). A pricing-page view might have a 14-day half-life (50% after 2 weeks). A whitepaper download might be 60 days.

Per-event decay is the most accurate model because different events represent different commitment levels and naturally age at different rates. It’s also harder to implement, which is why most systems default to group-level.

Why decay matters

Three problems decay solves:

1. Stale leads polluting “hot” lists. Without decay, a lead that did one high-value action a year ago shows the same score as one that did it last week. Sales loses trust in the threshold.
2. Inflation. All scoring models go up if there’s no downward pressure. Decay creates the gravity that keeps the distribution honest.
3. Recency signals. A burst of recent activity is meaningfully more predictive than steady accumulation. Decay weights recent events naturally higher.

Common decay failure modes

• No decay at all. Most damaging. Scores accumulate forever; the model rots quietly. This is the legacy HubSpot Score’s biggest weakness pre-Aug 2025.
• Same decay rate for all events. A pricing-page view and a demo request decaying at the same rate is a meaningful loss of resolution. Better than no decay, but not by as much as you’d hope.
• Decay too aggressive. If the half-life is too short, real engagement signals fall off before sales can act. If it’s too long, the decay is doing nothing.
• Decay with no reset. Some implementations decay continuously even during active engagement. Better to “refresh” decay timers when a related event fires.

Calibrating decay

The right decay rate is empirical. Pull closed-won deals, look at how long the engagement-to-close window was for each event type, and set the half-life around that window. A high-velocity SMB sale might have 30-day half-lives across the board; an enterprise sale with 9-month cycles might use 90- to 180-day half-lives.

Recalibrate quarterly. Decay rates that worked at Series A rarely work at Series C as the buyer profile shifts.

Related at kenbun

kenbun ships per-event half-life decay as a default. Each event type has its own decay curve, calibrated against your closed-won data when calibration tooling is enabled. The result: a pricing-page view from two weeks ago decays differently than a demo request from two months ago, which decays differently than a whitepaper download from six months ago. The model stays honest as time passes.