Definition
Recency-weighted engagement scoring is a methodology that assigns higher value to recent subscriber actions and proportionally less value to older actions, enabling more dynamic and responsive subscriber classification. Unlike simple binary classification (active if opened in 30 days, inactive if not), recency-weighted scoring produces a continuous engagement score that reflects both the presence and timing of engagement events. A subscriber who opened an email yesterday scores higher than one who opened 29 days ago, even though both fall within the 30-day active window.
The time-decay function determines how quickly older engagement loses value. Linear decay reduces the value of an engagement event by a fixed amount each day. A click 10 days ago might be worth 50% of a click today, and a click 20 days ago worth 25%. Exponential decay applies a constant percentage reduction per period, so each day the value of a past event is multiplied by a decay factor (e.g., 0.95). Exponential decay is more commonly used because it better reflects the pattern of subscriber engagement: the predictive value of a behaviour drops fastest immediately after it occurs and then plateaus. The decay rate should be calibrated using your own data to find the rate that best predicts future engagement.
Recency-weighted scoring produces three major benefits over simple classification. First, it enables more granular segmentation: instead of three engagement tiers (active, dormant, dead), you have a continuous score that can be divided into as many segments as operationally useful. Second, it enables earlier detection of engagement changes: a subscriber whose score drops 20 points in a week has changed behaviour, even if they remain in the "active" classification tier. Third, it improves targeting precision: campaigns can target specific score ranges rather than broad behavioural buckets, which typically improves conversion rates by 10-25% compared to simple segmentation.
Best Practices
Calibrate your decay function using predictive modelling on historical data. Split your subscriber base into a training dataset. For each subscriber, compute engagement scores using different decay functions and rates. Measure which function best predicts engagement in the next 30-day period. The optimal decay rate typically falls between 0.90 and 0.98 per day for most email programmes, meaning the value of an engagement event halves every 35-70 days.
Normalise recency-weighted scores for differences in send frequency. Subscribers who receive 10 emails per week have more opportunities to accumulate engagement score than those receiving 2 per week. Normalise scores by dividing by the number of engagement opportunities (emails received) in the scoring window. This prevents high-frequency sends from artificially inflating engagement scores.
Combine recency-weighted engagement with action-weighting. Not all actions are equal: a conversion should score higher than a click, which should score higher than an open. Apply action-weighting multipliers (e.g., conversion = 10 points, click = 3 points, open = 1 point) before applying the time-decay function. This creates a score that reflects both what the subscriber did and when they did it.
Use recency-weighted scores as inputs to AI-driven send-time optimisation and content selection. The continuous, granular nature of recency-weighted scores makes them ideal features for machine learning models. Send-time optimisation algorithms, content recommendation engines, and churn prediction models all perform better with recency-weighted inputs than with simple binary engagement flags.
Monitor score distribution shifts over time for early warning of list health changes. Plot the distribution of recency-weighted scores weekly. A leftward shift (more subscribers moving to lower scores) indicates a list-wide engagement decline that may need programmatic response. A stable distribution suggests the list is maintaining engagement levels.
Related Glossary Terms
Email Active Subscriber
An active email subscriber has opened or clicked an email within a defined recency period, typically 30-90 days by industry. Active subscriber rate of 40-60% is typical for healthy email lists.
Email Dormant Subscriber
A dormant subscriber has not engaged with emails for a defined inactivity period, typically 90-365 days depending on industry. Dormancy classification triggers suppression and re-engagement activities.
Email Engagement Tier
Email engagement tiering categorises subscribers into levels such as engaged, warming, passive, at-risk, dormant, and dead. Tier transition triggers enable automated send frequency and content strategy adjustments.
Email Frequency Engagement
Email frequency vs engagement analysis finds the optimal send frequency per subscriber. Over-sending increases unsubscribes and dormancy; under-sending diminishes brand recall. Testing methodology identifies the frequency sweet spot.
Email Inactivity Period
Email inactivity period measures the time since a subscriber last opened, clicked, or converted. Industry-specific benchmarks define thresholds for active, dormant, and at-risk classification based on inactivity windows.
Email Segment Overlap
Email segment overlap analysis identifies subscribers belonging to multiple segments. It enables cross-segment deduplication, priority rule management, and prevents campaign analytics double-counting.
Frequently Asked Questions
Simple classification uses a fixed recency threshold (e.g., active if opened in 30 days). Recency-weighted scoring produces a continuous score that reflects exactly when engagement occurred, enabling more precise targeting and earlier detection of behaviour changes. A subscriber who opened 5 days ago vs 25 days ago may have different scores even though both are classified as active.
Exponential decay is most commonly used because it reflects how the predictive value of past behaviour declines fastest immediately after the event. A decay factor of 0.93-0.97 per day (meaning the event value halves every 23-100 days) works well for most email programmes. The optimal rate depends on send frequency and purchase cycle.
Validate by measuring how well the score predicts future engagement. Split subscribers into score deciles and measure the actual engagement rate of each decile in the next 30-day period. A good model shows a monotonic relationship: each higher decile should have a progressively higher engagement rate. The ratio between the top and bottom decile engagement rates should be at least 5:1.
Yes. Spam complaints, hard bounces, and unsubscribe events can be recency-weighted with negative scores. A recent spam complaint should reduce the engagement score more heavily than an old one. This ensures that the scoring model is responsive to both positive and negative subscriber behaviour and adjusts treatment accordingly.
At least daily for any subscriber who receives email or takes engagement action. Real-time score updates are preferable for triggered campaigns where the score determines treatment. For batch campaigns, scores should be frozen at the time the campaign is queued to ensure consistent targeting across the send window.